[{"data":1,"prerenderedAt":1732},["ShallowReactive",2],{"blog-post-en-\u002Fblog\u002Fpaiton-qwen38-neo-gguf-vllm-r9700":3,"blog-posts-sidebar-en":1263},{"id":4,"title":5,"body":6,"categories":1246,"date":1251,"description":1252,"extension":1253,"heading":1254,"image":1255,"meta":1256,"navigation":695,"originalUrl":1257,"path":1258,"seo":1259,"slug":1260,"stem":1261,"__hash__":1262},"blog\u002Fblog\u002Fpaiton-qwen38-neo-gguf-vllm-r9700.md","Qwen3.8 GGUF in vLLM: Faster Responses on One Radeon",{"type":7,"value":8,"toc":1233},"minimark",[9,28,42,61,64,69,78,81,86,102,113,116,137,142,146,157,166,169,176,187,191,206,212,223,304,313,316,325,328,333,342,351,355,366,387,393,404,457,466,470,481,506,509,513,528,598,605,614,618,624,635,650,720,733,736,839,848,852,855,864,869,882,885,1229],[10,11,12,16],"p",{},[13,14,15],"strong",{},"Original NEO CODER MAX weights. Paiton-compiled execution. Up to 6.4% lower warm request latency than our matched llama.cpp baseline on a Radeon AI PRO R9700.",[17,18,19],"sup",{},[20,21,27],"a",{"href":22,"ariaDescribedBy":23,"dataFootnoteRef":25,"id":26},"#user-content-fn-readme",[24],"footnote-label","","user-content-fnref-readme","1",[10,29,30,34],{},[31,32,33],"em",{},"Original GGUF weights. Actual vLLM serving. Paiton-compiled execution on one Radeon AI PRO R9700.",[17,35,36],{},[20,37,41],{"href":38,"ariaDescribedBy":39,"dataFootnoteRef":25,"id":40},"#user-content-fn-native",[24],"user-content-fnref-native","2",[10,43,44,45,53],{},"DavidAU’s Qwen3.8 NEO CODER MAX has attracted plenty of attention. On September 14, 2026, its Hugging Face repository showed 875,703 downloads in the preceding month. That is a repository download count, not a count of unique users, but it is a useful indication of the interest around this fine-tune.",[17,46,47],{},[20,48,52],{"href":49,"ariaDescribedBy":50,"dataFootnoteRef":25,"id":51},"#user-content-fn-model",[24],"user-content-fnref-model","3",[17,54,55],{},[20,56,60],{"href":57,"ariaDescribedBy":58,"dataFootnoteRef":25,"id":59},"#user-content-fn-downloads",[24],"user-content-fnref-downloads","4",[10,62,63],{},"We wanted to answer a different question from another model review:",[10,65,66],{},[13,67,68],{},"Can you keep the original GGUF fine-tune, serve it through vLLM, and still compete with llama.cpp on a single Radeon?",[10,70,71,72],{},"For this tested model and configuration, yes. Paiton runs the selected Q4_K_M checkpoint through native AMD execution inside vLLM, rather than routing requests to a separate llama.cpp server.",[17,73,74],{},[20,75,41],{"href":38,"ariaDescribedBy":76,"dataFootnoteRef":25,"id":77},[24],"user-content-fnref-native-2",[10,79,80],{},"The fine-tune is the author’s. The execution path is ours.",[82,83,85],"h2",{"id":84},"gguf-support-is-not-the-same-as-optimized-gguf-execution","GGUF support is not the same as optimized GGUF execution",[10,87,88,89,93,94],{},"vLLM already supports GGUF. Its current documentation nevertheless describes that support as “highly experimental and under-optimized,” with potential incompatibilities with other features. Support now lives in the upstream ",[90,91,92],"code",{},"vllm-gguf-plugin",".",[17,95,96],{},[20,97,101],{"href":98,"ariaDescribedBy":99,"dataFootnoteRef":25,"id":100},"#user-content-fn-vllm",[24],"user-content-fnref-vllm","5",[10,103,104,105],{},"That plugin already documents several model families, including related Qwen vision-language models. This is not an announcement that GGUF has become possible in vLLM for the first time.",[17,106,107],{},[20,108,112],{"href":109,"ariaDescribedBy":110,"dataFootnoteRef":25,"id":111},"#user-content-fn-upstream",[24],"user-content-fnref-upstream","6",[10,114,115],{},"Our focus is narrower: make this particular GGUF fine-tune run efficiently on AMD hardware without making users switch to a different checkpoint or a different serving framework.",[10,117,118,119,127,128,136],{},"GGUF is a container for weights and metadata, not a requirement to use one inference engine.",[17,120,121],{},[20,122,126],{"href":123,"ariaDescribedBy":124,"dataFootnoteRef":25,"id":125},"#user-content-fn-gguf",[24],"user-content-fnref-gguf","7"," And llama.cpp already provides its own OpenAI-compatible server.",[17,129,130],{},[20,131,135],{"href":132,"ariaDescribedBy":133,"dataFootnoteRef":25,"id":134},"#user-content-fn-llama",[24],"user-content-fnref-llama","8"," An API wrapper alone would not be the achievement here.",[10,138,139],{},[13,140,141],{},"The distinction is actual vLLM serving, backed by Paiton’s compiled native execution.",[82,143,145],{"id":144},"keep-the-fine-tune-change-the-execution-path","Keep the fine-tune. Change the execution path.",[10,147,148,149],{},"The selected checkpoint is the author’s original mixed Q4_K_M GGUF, including its higher-precision tensors and BF16 output head. We have not replaced it with base Qwen or converted it into a new AWQ checkpoint.",[17,150,151],{},[20,152,156],{"href":153,"ariaDescribedBy":154,"dataFootnoteRef":25,"id":155},"#user-content-fn-lock",[24],"user-content-fnref-lock","9",[10,158,159,160],{},"Paiton provides the native language and vision execution. vLLM retains the model-loading integration, scheduling, sampling and streaming interface. The compiler remains proprietary; the public release contains the runtime artifacts needed to use it.",[17,161,162],{},[20,163,41],{"href":38,"ariaDescribedBy":164,"dataFootnoteRef":25,"id":165},[24],"user-content-fnref-native-3",[10,167,168],{},"This is an optimization inside an existing serving stack, not a requirement to adopt another inference server.",[10,170,171],{},[172,173],"img",{"alt":174,"src":175},"High-level integration: original pinned GGUF weights, actual vLLM scheduling, sampling and streaming, Paiton-compiled native execution, and one Radeon AI PRO R9700. The proprietary compiler is not distributed.","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-neo-gguf\u002F02-native-execution-neo-gguf.webp",[10,177,178,181],{},[31,179,180],{},"The public integration at a glance: keep the author’s GGUF, serve through vLLM, and execute with Paiton on the R9700. This is an integration overview, not a disclosure of compiler internals.",[17,182,183],{},[20,184,41],{"href":38,"ariaDescribedBy":185,"dataFootnoteRef":25,"id":186},[24],"user-content-fnref-native-4",[82,188,190],{"id":189},"faster-complete-responses-not-just-a-faster-kernel","Faster complete responses, not just a faster kernel",[10,192,193,194,197,198],{},"The comparison below measures ",[13,195,196],{},"complete streaming HTTP requests",", with exactly 128 generated tokens. Both engines ran serially on the same idle R9700. Each workload had one warmup and five measured requests.",[17,199,200],{},[20,201,205],{"href":202,"ariaDescribedBy":203,"dataFootnoteRef":25,"id":204},"#user-content-fn-bench",[24],"user-content-fnref-bench","10",[10,207,208],{},[172,209],{"alt":210,"src":211},"Median warm complete-request latency with 128 output tokens. At 128 input tokens, llama.cpp takes 5.264 seconds and Paiton plus vLLM 4.925 seconds, 6.4% lower. At 1,024 input, 5.933 versus 5.631 seconds, 5.1% lower. At 4,096 input, 9.099 versus 9.023 seconds, 0.8% lower, near-parity. One warmup and five measured requests per workload.","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-neo-gguf\u002F01-text-latency-neo-gguf.webp",[10,213,214,217],{},[31,215,216],{},"Complete streaming requests, not kernel-only timings. Both engines generate exactly 128 tokens. The 4,096-token result is near-parity; these five-request samples do not establish statistical significance.",[17,218,219],{},[20,220,205],{"href":202,"ariaDescribedBy":221,"dataFootnoteRef":25,"id":222},[24],"user-content-fnref-bench-2",[224,225,226,246],"table",{},[227,228,229],"thead",{},[230,231,232,236,240,243],"tr",{},[233,234,235],"th",{},"Input tokens",[233,237,239],{"align":238},"right","llama.cpp median",[233,241,242],{"align":238},"Paiton + vLLM median",[233,244,245],{"align":238},"Request latency reduction",[247,248,249,268,286],"tbody",{},[230,250,251,255,258,263],{},[252,253,254],"td",{},"128",[252,256,257],{"align":238},"5.264 s",[252,259,260],{"align":238},[13,261,262],{},"4.925 s",[252,264,265],{"align":238},[13,266,267],{},"6.4%",[230,269,270,273,276,281],{},[252,271,272],{},"1,024",[252,274,275],{"align":238},"5.933 s",[252,277,278],{"align":238},[13,279,280],{},"5.631 s",[252,282,283],{"align":238},[13,284,285],{},"5.1%",[230,287,288,291,294,299],{},[252,289,290],{},"4,096",[252,292,293],{"align":238},"9.099 s",[252,295,296],{"align":238},[13,297,298],{},"9.023 s",[252,300,301],{"align":238},[13,302,303],{},"0.8%",[10,305,306,307],{},"These are warm-request measurements, not first-install or cold-start timings.",[17,308,309],{},[20,310,205],{"href":202,"ariaDescribedBy":311,"dataFootnoteRef":25,"id":312},[24],"user-content-fnref-bench-3",[10,314,315],{},"The 4,096-token result is best described as near-parity. A 76-millisecond median difference in a small sample is not a basis for a sweeping performance claim.",[10,317,318,319],{},"Nor does lower complete-request latency mean every part of every request is faster: llama.cpp still wins some single-output, prefill-dominated tests.",[17,320,321],{},[20,322,41],{"href":38,"ariaDescribedBy":323,"dataFootnoteRef":25,"id":324},[24],"user-content-fnref-native-5",[10,326,327],{},"The useful result is that this GGUF can stay inside vLLM while delivering competitive latency. In these measured full-response cases, it was also lower.",[329,330,332],"h3",{"id":331},"what-was-held-constant","What was held constant?",[10,334,335,336],{},"The comparison used the same pinned GGUF, source tokenizer, 8,192-token context, 2,048-token prefill chunks and BF16 KV cache. Both engines had one active sequence, MTP and prefix caching disabled, and greedy sampling with fixed token counts. llama.cpp was an unmodified HIP build, not a CPU fallback.",[17,337,338],{},[20,339,205],{"href":202,"ariaDescribedBy":340,"dataFootnoteRef":25,"id":341},[24],"user-content-fnref-bench-4",[10,343,344,345],{},"The fixed-length prompts are synthetic timing workloads, not a coding-productivity benchmark. Shorter reasoning or earlier stopping is not being counted as faster execution.",[17,346,347],{},[20,348,205],{"href":202,"ariaDescribedBy":349,"dataFootnoteRef":25,"id":350},[24],"user-content-fnref-bench-5",[82,352,354],{"id":353},"image-input-works-too","Image input works too",[10,356,357,358],{},"This release also accepts one PNG or JPEG image through the chat-completions interface. Its image encoder runs through the native Paiton path; image embeddings and generated text share the same context budget.",[17,359,360],{},[20,361,365],{"href":362,"ariaDescribedBy":363,"dataFootnoteRef":25,"id":364},"#user-content-fn-image",[24],"user-content-fnref-image","11",[10,367,368,369,372,373,376,377,380,381],{},"For a ",[13,370,371],{},"1,024 × 1,024 image and 128 output tokens",", median complete-request time was ",[13,374,375],{},"6.279 seconds with Paiton versus 6.455 seconds with llama.cpp",", about ",[13,378,379],{},"2.7% lower",". That includes image processing, language prefill, generation and serving overhead.",[17,382,383],{},[20,384,41],{"href":38,"ariaDescribedBy":385,"dataFootnoteRef":25,"id":386},[24],"user-content-fnref-native-6",[10,388,389],{},[172,390],{"alt":391,"src":392},"Median complete-request latency for a single image and 128 output tokens. A 256 by 256 image takes 5.111 seconds in llama.cpp versus 4.909 seconds in Paiton plus vLLM, 4.0% lower. A 1,024 by 1,024 image takes 6.455 versus 6.279 seconds, 2.7% lower. Five measured requests per case.","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-neo-gguf\u002F03-image-latency-neo-gguf.webp",[10,394,395,398],{},[31,396,397],{},"Two qualified image sizes, each followed by 128 generated tokens. Timing includes image encoding, language prefill, generation and serving overhead. This measures image understanding, not image generation.",[17,399,400],{},[20,401,205],{"href":202,"ariaDescribedBy":402,"dataFootnoteRef":25,"id":403},[24],"user-content-fnref-bench-6",[224,405,406,419],{},[227,407,408],{},[230,409,410,413,415,417],{},[233,411,412],{},"Image size",[233,414,239],{"align":238},[233,416,242],{"align":238},[233,418,245],{"align":238},[247,420,421,439],{},[230,422,423,426,429,434],{},[252,424,425],{},"256 × 256",[252,427,428],{"align":238},"5.111 s",[252,430,431],{"align":238},[13,432,433],{},"4.909 s",[252,435,436],{"align":238},[13,437,438],{},"4.0%",[230,440,441,444,447,452],{},[252,442,443],{},"1,024 × 1,024",[252,445,446],{"align":238},"6.455 s",[252,448,449],{"align":238},[13,450,451],{},"6.279 s",[252,453,454],{"align":238},[13,455,456],{},"2.7%",[10,458,459,460],{},"This is single-image inference, not video or unrestricted multi-image support.",[17,461,462],{},[20,463,365],{"href":362,"ariaDescribedBy":464,"dataFootnoteRef":25,"id":465},[24],"user-content-fnref-image-2",[82,467,469],{"id":468},"same-weights-does-not-mean-identical-arithmetic","Same weights does not mean identical arithmetic",[10,471,472,473],{},"Keeping a GGUF checkpoint unchanged does not make two runtimes numerically identical. The qualified execution profile uses different activation arithmetic from the older FP32 reference path; the weight values remain unchanged.",[17,474,475],{},[20,476,480],{"href":477,"ariaDescribedBy":478,"dataFootnoteRef":25,"id":479},"#user-content-fn-release",[24],"user-content-fnref-release","12",[10,482,483,484,487,488,491,492,498],{},"The published checks include ",[13,485,486],{},"10 of 11 fixed text tasks passed",", with the same failed task as llama.cpp, and ",[13,489,490],{},"five image fixtures passed",". The final prefill optimization also matched 31,784,960 compared logits against the preceding qualified Paiton profile, not against every other engine.",[17,493,494],{},[20,495,480],{"href":477,"ariaDescribedBy":496,"dataFootnoteRef":25,"id":497},[24],"user-content-fnref-release-2",[17,499,500],{},[20,501,505],{"href":502,"ariaDescribedBy":503,"dataFootnoteRef":25,"id":504},"#user-content-fn-publication",[24],"user-content-fnref-publication","13",[10,507,508],{},"Those are useful release checks, not proof of unchanged capability on every coding task, conversation or image. We keep the numerical comparisons and their reference profiles explicit rather than describing the release as universally bit-identical.",[82,510,512],{"id":511},"the-deployment-profile","The deployment profile",[10,514,515,516,522],{},"The published v1.1.0 release is qualified for the following configuration:",[17,517,518],{},[20,519,480],{"href":477,"ariaDescribedBy":520,"dataFootnoteRef":25,"id":521},[24],"user-content-fnref-release-3",[17,523,524],{},[20,525,365],{"href":362,"ariaDescribedBy":526,"dataFootnoteRef":25,"id":527},[24],"user-content-fnref-image-3",[224,529,530,540],{},[227,531,532],{},[230,533,534,537],{},[233,535,536],{},"Setting",[233,538,539],{},"Supported profile",[247,541,542,550,558,566,574,582,590],{},[230,543,544,547],{},[252,545,546],{},"GPU",[252,548,549],{},"One Radeon AI PRO R9700, gfx1201",[230,551,552,555],{},[252,553,554],{},"Model",[252,556,557],{},"Pinned original NEO CODER MAX Q4_K_M GGUF",[230,559,560,563],{},[252,561,562],{},"Runtime",[252,564,565],{},"Paiton; pinned ROCm 7.14.60850",[230,567,568,571],{},[252,569,570],{},"Context",[252,572,573],{},"8,192 tokens total",[230,575,576,579],{},[252,577,578],{},"Active sequences",[252,580,581],{},"One; additional HTTP requests queue",[230,583,584,587],{},[252,585,586],{},"Input",[252,588,589],{},"Text, or text with one PNG\u002FJPEG image",[230,591,592,595],{},[252,593,594],{},"MTP, prefix caching, video",[252,596,597],{},"Disabled",[10,599,600,601,604],{},"The model’s name includes MTP, but these results do ",[13,602,603],{},"not"," use speculative MTP decoding. Likewise, queued clients should not be confused with qualified multi-sequence GPU batching.",[10,606,607,608],{},"This release targets the 32 GB R9700. It is not a promise of support for a smaller GPU, another GGUF quantization, longer contexts or high-concurrency serving.",[17,609,610],{},[20,611,27],{"href":22,"ariaDescribedBy":612,"dataFootnoteRef":25,"id":613},[24],"user-content-fnref-readme-2",[82,615,617],{"id":616},"run-the-published-release","Run the published release",[10,619,620],{},[172,621],{"alt":622,"src":623},"The qualified local profile: one Radeon AI PRO R9700, Linux and Docker, an 8,192-token context, one active sequence, and text or one PNG\u002FJPEG image. An illustrative API view shows the local chat-completions endpoint and qwen38-neo model name. Not an application screenshot.","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-neo-gguf\u002F05-local-endpoint-neo-gguf.webp",[10,625,626,629],{},[31,627,628],{},"The qualified local serving profile. The API panel is illustrative, not an application screenshot or benchmark trace. Compiler access is not needed to use the published runtime.",[17,630,631],{},[20,632,480],{"href":477,"ariaDescribedBy":633,"dataFootnoteRef":25,"id":634},[24],"user-content-fnref-release-4",[10,636,637,638,644],{},"On a Linux system with an R9700, a working AMD driver and Docker GPU access, clone the public helper repository and launch the pinned v1.1.0 image:",[17,639,640],{},[20,641,27],{"href":22,"ariaDescribedBy":642,"dataFootnoteRef":25,"id":643},[24],"user-content-fnref-readme-3",[17,645,646],{},[20,647,480],{"href":477,"ariaDescribedBy":648,"dataFootnoteRef":25,"id":649},[24],"user-content-fnref-release-5",[651,652,656],"pre",{"className":653,"code":654,"language":655,"meta":25,"style":25},"language-bash shiki shiki-themes github-light github-dark","git clone --depth 1 https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin.git\ncd paiton-vllm-plugin\n\nPAITON_NEO_IMAGE=ghcr.io\u002Feliovp\u002Fpaiton-vllm-plugin@sha256:534287969135f581744ae481b578599468b0bf7ac9a4051b0941500e4c18da4d \\\n  .\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002Fserve-docker.sh\n","bash",[90,657,658,681,690,697,714],{"__ignoreMap":25},[659,660,663,667,671,675,678],"span",{"class":661,"line":662},"line",1,[659,664,666],{"class":665},"sScJk","git",[659,668,670],{"class":669},"sZZnC"," clone",[659,672,674],{"class":673},"sj4cs"," --depth",[659,676,677],{"class":673}," 1",[659,679,680],{"class":669}," https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin.git\n",[659,682,684,687],{"class":661,"line":683},2,[659,685,686],{"class":673},"cd",[659,688,689],{"class":669}," paiton-vllm-plugin\n",[659,691,693],{"class":661,"line":692},3,[659,694,696],{"emptyLinePlaceholder":695},true,"\n",[659,698,700,704,708,711],{"class":661,"line":699},4,[659,701,703],{"class":702},"sVt8B","PAITON_NEO_IMAGE",[659,705,707],{"class":706},"szBVR","=",[659,709,710],{"class":669},"ghcr.io\u002Feliovp\u002Fpaiton-vllm-plugin@sha256:534287969135f581744ae481b578599468b0bf7ac9a4051b0941500e4c18da4d",[659,712,713],{"class":665}," \\\n",[659,715,717],{"class":661,"line":716},5,[659,718,719],{"class":669},"  .\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002Fserve-docker.sh\n",[10,721,722,723,726,727],{},"The first start downloads approximately ",[13,724,725],{},"19.43 GB"," of pinned language and projector files. Later starts reuse the cache and verify the hashes. No compiler checkout is required. An existing copy of the exact pinned GGUF can also be mounted and reused using the model guide’s instructions.",[17,728,729],{},[20,730,27],{"href":22,"ariaDescribedBy":731,"dataFootnoteRef":25,"id":732},[24],"user-content-fnref-readme-4",[10,734,735],{},"After the server is ready, send a request from another terminal:",[651,737,739],{"className":653,"code":738,"language":655,"meta":25,"style":25},"curl http:\u002F\u002F127.0.0.1:8000\u002Fv1\u002Fchat\u002Fcompletions \\\n  -H 'Content-Type: application\u002Fjson' \\\n  -d '{\n    \"model\": \"qwen38-neo\",\n    \"messages\": [\n      {\n        \"role\": \"user\",\n        \"content\": \"Write a Python function that removes duplicate integers while preserving their original order.\"\n      }\n    ],\n    \"temperature\": 0,\n    \"max_tokens\": 256,\n    \"stream\": true,\n    \"chat_template_kwargs\": {\"enable_thinking\": false}\n  }'\n",[90,740,741,751,761,769,774,779,785,791,797,803,809,815,821,827,833],{"__ignoreMap":25},[659,742,743,746,749],{"class":661,"line":662},[659,744,745],{"class":665},"curl",[659,747,748],{"class":669}," http:\u002F\u002F127.0.0.1:8000\u002Fv1\u002Fchat\u002Fcompletions",[659,750,713],{"class":673},[659,752,753,756,759],{"class":661,"line":683},[659,754,755],{"class":673},"  -H",[659,757,758],{"class":669}," 'Content-Type: application\u002Fjson'",[659,760,713],{"class":673},[659,762,763,766],{"class":661,"line":692},[659,764,765],{"class":673},"  -d",[659,767,768],{"class":669}," '{\n",[659,770,771],{"class":661,"line":699},[659,772,773],{"class":669},"    \"model\": \"qwen38-neo\",\n",[659,775,776],{"class":661,"line":716},[659,777,778],{"class":669},"    \"messages\": [\n",[659,780,782],{"class":661,"line":781},6,[659,783,784],{"class":669},"      {\n",[659,786,788],{"class":661,"line":787},7,[659,789,790],{"class":669},"        \"role\": \"user\",\n",[659,792,794],{"class":661,"line":793},8,[659,795,796],{"class":669},"        \"content\": \"Write a Python function that removes duplicate integers while preserving their original order.\"\n",[659,798,800],{"class":661,"line":799},9,[659,801,802],{"class":669},"      }\n",[659,804,806],{"class":661,"line":805},10,[659,807,808],{"class":669},"    ],\n",[659,810,812],{"class":661,"line":811},11,[659,813,814],{"class":669},"    \"temperature\": 0,\n",[659,816,818],{"class":661,"line":817},12,[659,819,820],{"class":669},"    \"max_tokens\": 256,\n",[659,822,824],{"class":661,"line":823},13,[659,825,826],{"class":669},"    \"stream\": true,\n",[659,828,830],{"class":661,"line":829},14,[659,831,832],{"class":669},"    \"chat_template_kwargs\": {\"enable_thinking\": false}\n",[659,834,836],{"class":661,"line":835},15,[659,837,838],{"class":669},"  }'\n",[10,840,841,842],{},"The example explicitly disables thinking through the retained source template. It is an interactive example, not the fixed 128-output-token benchmark request.",[17,843,844],{},[20,845,27],{"href":22,"ariaDescribedBy":846,"dataFootnoteRef":25,"id":847},[24],"user-content-fnref-readme-5",[82,849,851],{"id":850},"more-useful-performance-without-another-serving-stack","More useful performance, without another serving stack",[10,853,854],{},"The point is not that every model needs a new engine. It is that a useful fine-tune should not need to lose its identity to fit an optimized deployment path.",[10,856,857,858],{},"For this release, the original GGUF and vLLM stay in place. Paiton changes the execution underneath them.",[17,859,860],{},[20,861,41],{"href":38,"ariaDescribedBy":862,"dataFootnoteRef":25,"id":863},[24],"user-content-fnref-native-7",[10,865,866],{},[13,867,868],{},"Keep the fine-tune. Keep vLLM. Get more from the Radeon you already have.",[10,870,871,872,93,876],{},"Start with the public model guide and its benchmark report. For AMD inference workloads beyond this qualified profile, talk to us about ",[20,873,875],{"href":874},"\u002Fproducts\u002Fpaiton","Paiton",[17,877,878],{},[20,879,27],{"href":22,"ariaDescribedBy":880,"dataFootnoteRef":25,"id":881},[24],"user-content-fnref-readme-6",[883,884],"hr",{},[886,887,890,895],"section",{"className":888,"dataFootnotes":25},[889],"footnotes",[82,891,894],{"className":892,"id":24},[893],"sr-only","Footnotes",[896,897,898,952,1007,1021,1034,1048,1061,1074,1087,1100,1148,1175,1216],"ol",{},[899,900,902,908,909,916,917,916,924,916,931,916,938,916,945],"li",{"id":901},"user-content-fn-readme",[20,903,907],{"href":904,"rel":905},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Fblob\u002Fmain\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002FREADME.md",[906],"nofollow","Paiton NEO model guide",". ",[20,910,915],{"href":911,"ariaLabel":912,"className":913,"dataFootnoteBackref":25},"#user-content-fnref-readme","Back to reference 1",[914],"data-footnote-backref","↩"," ",[20,918,915,922],{"href":919,"ariaLabel":920,"className":921,"dataFootnoteBackref":25},"#user-content-fnref-readme-2","Back to reference 1-2",[914],[17,923,41],{},[20,925,915,929],{"href":926,"ariaLabel":927,"className":928,"dataFootnoteBackref":25},"#user-content-fnref-readme-3","Back to reference 1-3",[914],[17,930,52],{},[20,932,915,936],{"href":933,"ariaLabel":934,"className":935,"dataFootnoteBackref":25},"#user-content-fnref-readme-4","Back to reference 1-4",[914],[17,937,60],{},[20,939,915,943],{"href":940,"ariaLabel":941,"className":942,"dataFootnoteBackref":25},"#user-content-fnref-readme-5","Back to reference 1-5",[914],[17,944,101],{},[20,946,915,950],{"href":947,"ariaLabel":948,"className":949,"dataFootnoteBackref":25},"#user-content-fnref-readme-6","Back to reference 1-6",[914],[17,951,112],{},[899,953,955,908,960,916,965,916,972,916,979,916,986,916,993,916,1000],{"id":954},"user-content-fn-native",[20,956,959],{"href":957,"rel":958},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Fblob\u002Fmain\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002FNATIVE_GGUF.md",[906],"Native GGUF through vLLM on AMD RDNA4",[20,961,915],{"href":962,"ariaLabel":963,"className":964,"dataFootnoteBackref":25},"#user-content-fnref-native","Back to reference 2",[914],[20,966,915,970],{"href":967,"ariaLabel":968,"className":969,"dataFootnoteBackref":25},"#user-content-fnref-native-2","Back to reference 2-2",[914],[17,971,41],{},[20,973,915,977],{"href":974,"ariaLabel":975,"className":976,"dataFootnoteBackref":25},"#user-content-fnref-native-3","Back to reference 2-3",[914],[17,978,52],{},[20,980,915,984],{"href":981,"ariaLabel":982,"className":983,"dataFootnoteBackref":25},"#user-content-fnref-native-4","Back to reference 2-4",[914],[17,985,60],{},[20,987,915,991],{"href":988,"ariaLabel":989,"className":990,"dataFootnoteBackref":25},"#user-content-fnref-native-5","Back to reference 2-5",[914],[17,992,101],{},[20,994,915,998],{"href":995,"ariaLabel":996,"className":997,"dataFootnoteBackref":25},"#user-content-fnref-native-6","Back to reference 2-6",[914],[17,999,112],{},[20,1001,915,1005],{"href":1002,"ariaLabel":1003,"className":1004,"dataFootnoteBackref":25},"#user-content-fnref-native-7","Back to reference 2-7",[914],[17,1006,126],{},[899,1008,1010,1015,1016],{"id":1009},"user-content-fn-model",[20,1011,1014],{"href":1012,"rel":1013},"https:\u002F\u002Fhuggingface.co\u002FDavidAU\u002FQwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NEO-CODER-MAX-MTP-GGUF",[906],"DavidAU model card",", download snapshot checked September 14, 2026. ",[20,1017,915],{"href":1018,"ariaLabel":1019,"className":1020,"dataFootnoteBackref":25},"#user-content-fnref-model","Back to reference 3",[914],[899,1022,1024,908,1029],{"id":1023},"user-content-fn-downloads",[20,1025,1028],{"href":1026,"rel":1027},"https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Fhub\u002Fmodels-download-stats",[906],"Hugging Face: Models Download Stats",[20,1030,915],{"href":1031,"ariaLabel":1032,"className":1033,"dataFootnoteBackref":25},"#user-content-fnref-downloads","Back to reference 4",[914],[899,1035,1037,1042,1043],{"id":1036},"user-content-fn-vllm",[20,1038,1041],{"href":1039,"rel":1040},"https:\u002F\u002Fdocs.vllm.ai\u002Fen\u002Fstable\u002Ffeatures\u002Fquantization\u002Fgguf\u002F",[906],"vLLM: GGUF documentation",", checked September 14, 2026. ",[20,1044,915],{"href":1045,"ariaLabel":1046,"className":1047,"dataFootnoteBackref":25},"#user-content-fnref-vllm","Back to reference 5",[914],[899,1049,1051,908,1056],{"id":1050},"user-content-fn-upstream",[20,1052,1055],{"href":1053,"rel":1054},"https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm-gguf-plugin",[906],"Upstream vllm-gguf-plugin",[20,1057,915],{"href":1058,"ariaLabel":1059,"className":1060,"dataFootnoteBackref":25},"#user-content-fnref-upstream","Back to reference 6",[914],[899,1062,1064,908,1069],{"id":1063},"user-content-fn-gguf",[20,1065,1068],{"href":1066,"rel":1067},"https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Fhub\u002Fgguf",[906],"Hugging Face: GGUF",[20,1070,915],{"href":1071,"ariaLabel":1072,"className":1073,"dataFootnoteBackref":25},"#user-content-fnref-gguf","Back to reference 7",[914],[899,1075,1077,908,1082],{"id":1076},"user-content-fn-llama",[20,1078,1081],{"href":1079,"rel":1080},"https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp",[906],"llama.cpp",[20,1083,915],{"href":1084,"ariaLabel":1085,"className":1086,"dataFootnoteBackref":25},"#user-content-fnref-llama","Back to reference 8",[914],[899,1088,1090,908,1095],{"id":1089},"user-content-fn-lock",[20,1091,1094],{"href":1092,"rel":1093},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Fblob\u002Fmain\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002Fcheckpoint.lock.json",[906],"Checkpoint lock",[20,1096,915],{"href":1097,"ariaLabel":1098,"className":1099,"dataFootnoteBackref":25},"#user-content-fnref-lock","Back to reference 9",[914],[899,1101,1103,908,1108,916,1113,916,1120,916,1127,916,1134,916,1141],{"id":1102},"user-content-fn-bench",[20,1104,1107],{"href":1105,"rel":1106},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Fblob\u002Fmain\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002FBENCHMARKS.md",[906],"NEO v1.1.0 benchmarks",[20,1109,915],{"href":1110,"ariaLabel":1111,"className":1112,"dataFootnoteBackref":25},"#user-content-fnref-bench","Back to reference 10",[914],[20,1114,915,1118],{"href":1115,"ariaLabel":1116,"className":1117,"dataFootnoteBackref":25},"#user-content-fnref-bench-2","Back to reference 10-2",[914],[17,1119,41],{},[20,1121,915,1125],{"href":1122,"ariaLabel":1123,"className":1124,"dataFootnoteBackref":25},"#user-content-fnref-bench-3","Back to reference 10-3",[914],[17,1126,52],{},[20,1128,915,1132],{"href":1129,"ariaLabel":1130,"className":1131,"dataFootnoteBackref":25},"#user-content-fnref-bench-4","Back to reference 10-4",[914],[17,1133,60],{},[20,1135,915,1139],{"href":1136,"ariaLabel":1137,"className":1138,"dataFootnoteBackref":25},"#user-content-fnref-bench-5","Back to reference 10-5",[914],[17,1140,101],{},[20,1142,915,1146],{"href":1143,"ariaLabel":1144,"className":1145,"dataFootnoteBackref":25},"#user-content-fnref-bench-6","Back to reference 10-6",[914],[17,1147,112],{},[899,1149,1151,908,1156,916,1161,916,1168],{"id":1150},"user-content-fn-image",[20,1152,1155],{"href":1153,"rel":1154},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Fblob\u002Fmain\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002FIMAGE_API.md",[906],"Image API",[20,1157,915],{"href":1158,"ariaLabel":1159,"className":1160,"dataFootnoteBackref":25},"#user-content-fnref-image","Back to reference 11",[914],[20,1162,915,1166],{"href":1163,"ariaLabel":1164,"className":1165,"dataFootnoteBackref":25},"#user-content-fnref-image-2","Back to reference 11-2",[914],[17,1167,41],{},[20,1169,915,1173],{"href":1170,"ariaLabel":1171,"className":1172,"dataFootnoteBackref":25},"#user-content-fnref-image-3","Back to reference 11-3",[914],[17,1174,52],{},[899,1176,1178,908,1183,916,1188,916,1195,916,1202,916,1209],{"id":1177},"user-content-fn-release",[20,1179,1182],{"href":1180,"rel":1181},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Fblob\u002Fmain\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002Fpaiton-release.json",[906],"v1.1.0 release manifest",[20,1184,915],{"href":1185,"ariaLabel":1186,"className":1187,"dataFootnoteBackref":25},"#user-content-fnref-release","Back to reference 12",[914],[20,1189,915,1193],{"href":1190,"ariaLabel":1191,"className":1192,"dataFootnoteBackref":25},"#user-content-fnref-release-2","Back to reference 12-2",[914],[17,1194,41],{},[20,1196,915,1200],{"href":1197,"ariaLabel":1198,"className":1199,"dataFootnoteBackref":25},"#user-content-fnref-release-3","Back to reference 12-3",[914],[17,1201,52],{},[20,1203,915,1207],{"href":1204,"ariaLabel":1205,"className":1206,"dataFootnoteBackref":25},"#user-content-fnref-release-4","Back to reference 12-4",[914],[17,1208,60],{},[20,1210,915,1214],{"href":1211,"ariaLabel":1212,"className":1213,"dataFootnoteBackref":25},"#user-content-fnref-release-5","Back to reference 12-5",[914],[17,1215,101],{},[899,1217,1219,908,1224],{"id":1218},"user-content-fn-publication",[20,1220,1223],{"href":1221,"rel":1222},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Fblob\u002Fmain\u002Fmodels\u002FQwen3.8-NEO-CODER-MAX\u002Fpublication-checks.json",[906],"Publication checks",[20,1225,915],{"href":1226,"ariaLabel":1227,"className":1228,"dataFootnoteBackref":25},"#user-content-fnref-publication","Back to reference 13",[914],[1230,1231,1232],"style",{},"html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}html pre.shiki code .szBVR, html code.shiki .szBVR{--shiki-default:#D73A49;--shiki-dark:#F97583}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"title":25,"searchDepth":683,"depth":683,"links":1234},[1235,1236,1237,1240,1241,1242,1243,1244,1245],{"id":84,"depth":683,"text":85},{"id":144,"depth":683,"text":145},{"id":189,"depth":683,"text":190,"children":1238},[1239],{"id":331,"depth":692,"text":332},{"id":353,"depth":683,"text":354},{"id":468,"depth":683,"text":469},{"id":511,"depth":683,"text":512},{"id":616,"depth":683,"text":617},{"id":850,"depth":683,"text":851},{"id":24,"depth":683,"text":894},[875,1247,1248,1249,1250],"AMD Radeon","Local AI","GGUF","vLLM","2026-09-14T07:30:00Z","Run the original NEO CODER MAX GGUF in vLLM with Paiton on an R9700. Explore measured latency gains, image input and local deployment.","md",null,"\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-neo-gguf\u002F00-hero-neo-gguf-r9700.webp",{},"https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-neo-gguf-vllm-r9700","\u002Fblog\u002Fpaiton-qwen38-neo-gguf-vllm-r9700",{"title":5,"description":1252},"paiton-qwen38-neo-gguf-vllm-r9700","blog\u002Fpaiton-qwen38-neo-gguf-vllm-r9700","o4h8PDZjhSKb__BYhF9Ga7Cke_lSVhFMKoWiJzLfmuY",[1264,1266,1278,1288,1304,1313,1327,1358,1370,1393,1411,1430,1448,1466,1483,1495,1511,1526,1538,1547,1555,1570,1582,1593,1604,1614,1627,1637,1650,1661,1671,1682,1691,1703,1714,1723],{"path":1258,"title":5,"description":1252,"date":1251,"slug":1260,"image":1255,"originalUrl":1257,"categories":1265},[875,1247,1248,1249,1250],{"path":1267,"title":1268,"description":1269,"date":1270,"slug":1271,"image":1272,"originalUrl":1273,"categories":1274},"\u002Fblog\u002Fpaiton-minimax-h3-radeon-ai-pro-r9700","MiniMax H3 on Radeon: 15-Second Video With Native Sound","Paiton generates a 15-second MiniMax H3 video with stereo audio on one Radeon AI PRO R9700 in 5m 33s, with 16.7% lower latency than matched stock.","2026-09-09T07:30:00Z","paiton-minimax-h3-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-minimax-h3\u002F00-featured-minimax-h3-r9700.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-minimax-h3-radeon-ai-pro-r9700",[875,1247,1248,1275,1276,1277],"Video Generation","MiniMax H3","ComfyUI",{"path":1279,"title":1280,"description":1281,"date":1282,"slug":1283,"image":1284,"originalUrl":1279,"categories":1285},"\u002Fblog\u002Fpaiton-flux2-klein-radeon-ai-pro-r9700","Local FLUX.2 klein on Radeon AI PRO R9700: Faster Image Generation with Less VRAM","Paiton generates 1024×1024 FLUX.2 klein images in 1.054 seconds on an R9700: 16.2% lower warm latency and 33.4% lower peak Torch allocation.","2026-09-07T09:00:00","paiton-flux2-klein-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-flux2-klein\u002Ffox-paiton.webp",[875,1247,1248,1286,1287,1277],"Image Generation","FLUX",{"path":1289,"title":1290,"description":1291,"date":1292,"slug":1293,"image":1294,"originalUrl":1295,"categories":1296},"\u002Fblog\u002Fpaiton-ornith15-radeon-ai-pro-r9700","Ornith 1.5 at 44.6 tok\u002Fs on One Radeon AI PRO R9700","Paiton serves Ornith 1.5 35B A3B at 44.63 output tok\u002Fs on one Radeon AI PRO R9700, 27% faster than tuned stock vLLM, with 21.3% lower modeled cost.","2026-09-05T09:00:00","paiton-ornith15-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-ornith15\u002F00-featured-ornith15-r9700.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-ornith15-radeon-ai-pro-r9700",[875,1297,1247,1298,1299,1300,1301,1302,1250,1303],"Artificial Intelligence","AI Inference","GPU Performance","Inference Latency","Inference Optimization","Large Language Models","Cost Efficiency",{"path":1305,"title":1306,"description":1307,"date":1308,"slug":1309,"image":1310,"originalUrl":1311,"categories":1312},"\u002Fblog\u002Fpaiton-qwen38-radeon-ai-pro-r9700","Paiton: 21% More Qwen3.8 Throughput on Radeon AI PRO R9700","Paiton serves AMD’s Qwen3.8 27B at 39.77 output tokens\u002Fs on one Radeon AI PRO R9700, delivering 21% more throughput and 17.4% lower modeled cost.","2026-09-04T09:00:00","paiton-qwen38-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-r9700\u002F00-featured-paiton-r9700.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-radeon-ai-pro-r9700",[875,1297,1247,1298,1299,1300,1301,1302,1250,1303],{"path":1314,"title":1315,"description":1316,"date":1317,"slug":1318,"image":1319,"originalUrl":1254,"categories":1320},"\u002Fblog\u002Fai-data-center-power-requirements-gpu-per-megawatt","AI Data Center Power Requirements: The GPU\u002FMW Illusion","Why do AI data-center proposals quote different GPU capacities? See how PUE, peak loads, storage, networking and cooling determine deployable compute.","2026-07-27T23:52:00","ai-data-center-power-requirements-gpu-per-megawatt","\u002Fasset\u002Fimages\u002Fblog\u002Fai-data-center-power-requirements-gpu-per-megawatt\u002Fgpu-per-megawatt-illusion.webp",[1321,1322,1323,1324,1325,1326],"All","AI Infrastructure","Data Centers","ModFlex","HPC","AMD Helios",{"path":1328,"title":1329,"description":1330,"date":1331,"slug":1332,"image":1333,"originalUrl":1334,"categories":1335},"\u002Fblog\u002Fpaiton-returns-to-its-diffusion-roots-optimizing-wan2-2-t2v-a14b-on-amd-mi355x","Paiton Returns to Its Diffusion Roots: Optimizing Wan2.2-T2V-A14B on AMD MI355X","When we first started building Paiton, one of our earliest focus areas was optimizing diffusion models. Stable Diffusion XL was one of the first large models where we showed that fused operators, efficient execution, and hardware-aware kernels could make a real difference. Now we are returning to those origins.With the growing interest in text-to-video generation, ...","2026-06-10T14:04:04","paiton-returns-to-its-diffusion-roots-optimizing-wan2-2-t2v-a14b-on-amd-mi355x","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaitonwan2.webp","https:\u002F\u002Feliovp.com\u002Fpaiton-returns-to-its-diffusion-roots-optimizing-wan2-2-t2v-a14b-on-amd-mi355x\u002F",[1321,1297,875,1336,1337,1338,1339,1340,1341,1342,1343,1344,546,1345,1346,1347,1348,1349,1350,1351,875,1352,1353,1354,1355,1356,1357],"14B","AMD","B200","Benchmarks","Blackwell","Compute","Diffusion","Eliovp","Generative AI","Hardware","Inference","Instinct","MI355x","NVidia","On-Premise","Optimization","Sovereign AI","T2V","Text-to-Video","Tuning","Video-Generation","Wan2.2",{"path":1359,"title":1360,"description":1361,"date":1362,"slug":1363,"image":1364,"originalUrl":1365,"categories":1366},"\u002Fblog\u002Ffrom-the-attic-to-the-front-page-eliovp-recognized-as-a-pioneer-in-chip-optimization-data-center-infrastructure","From the Attic to the Front Page: ElioVP Recognized as a Pioneer in Chip Optimization & Data Center Infrastructure","It has been some incredible weeks for the team here at Eliovp. We are extremely proud to share that our company was recently featured on the front page of De Tijd, Belgium’s leading business newspaper. Seeing our story, from our founder’s early days tinkering with wires in an attic to generating €215 million in revenue, ...","2026-02-10T20:48:12","from-the-attic-to-the-front-page-eliovp-recognized-as-a-pioneer-in-chip-optimization-data-center-infrastructure","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fphysicalnewspaper.webp","https:\u002F\u002Feliovp.com\u002Ffrom-the-attic-to-the-front-page-eliovp-recognized-as-a-pioneer-in-chip-optimization-data-center-infrastructure\u002F",[1321,1297,1367,1368,1337,1369,1367,1349],"Modular DC","Uncategorized","De Tijd",{"path":1371,"title":1372,"description":1373,"date":1374,"slug":1375,"image":1376,"originalUrl":1377,"categories":1378},"\u002Fblog\u002Fprivacy-is-geen-it-probleem-meer-het-is-een-strategische-prioriteit","Privacy is geen IT-probleem meer, het is een strategische prioriteit: ","Privacyrisico’s, Psychologische Valkuilen en de Operationele Realiteit van Generatieve AI in de Benelux De recente verschijning van een frontpage-artikel over ons bedrijf in het gerespecteerde dagblad “De Tijd” heeft onze zichtbaarheid aanzienlijk vergroot, wat de aanleiding is voor dit artikel. Deze mediabelangstelling, gecombineerd met de talrijke uitnodigingen voor spreekbeurten die we hebben ontvangen, fungeert als ...","2026-01-29T13:51:11","privacy-is-geen-it-probleem-meer-het-is-een-strategische-prioriteit","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fheaderimage.webp","https:\u002F\u002Feliovp.com\u002Fprivacy-is-geen-it-probleem-meer-het-is-een-strategische-prioriteit\u002F",[1321,1297,1379,1368,1380,1381,1382,1383,1384,1385,1386,1387,1343,1388,1389,1390,1391,1392],"Trending","AI Act","Antropomorfisme","AVG","Benelux","ChatGPT","Cybersecurity","Data Governance","Data Security","GDPR","Generatieve AI","Microsoft Copilot","Privacy","Shadow AI",{"path":1394,"title":1395,"description":1396,"date":1397,"slug":1398,"image":1399,"originalUrl":1400,"categories":1401},"\u002Fblog\u002Fitsme-bij-ons-is-het-its-not-me-en-dit-is-waarom","Itsme? Bij ons is het “it’s not me”, en dit is waarom.","1. Inleiding: De Strategische Noodzaak van Weigering In het hedendaagse digitale landschap wordt de keuze voor een identiteit leverancier (IdP) vaak gereduceerd tot een discussie over User Experience (UX) en conversieratio’s. Deze reductionistische benadering verhult echter de diepgaande strategische, juridische en operationele risico’s die gepaard gaan met het uitbesteden van de “Sleutels tot het Koninkrijk”, ...","2025-11-27T09:32:14","itsme-bij-ons-is-het-its-not-me-en-dit-is-waarom","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ffrontimage.webp","https:\u002F\u002Feliovp.com\u002Fitsme-bij-ons-is-het-its-not-me-en-dit-is-waarom\u002F",[1321,1402,1403,1404,1385,1405,1406,1407,1388,1408,1409,1410,1391],"AWS","Belgian Mobile ID","Cloud Act","Data Soevereiniteit","Digitale Identiteit","eIDAS","itsme","Liberty Global","MyGov.be",{"path":1412,"title":1413,"description":1414,"date":1415,"slug":1416,"image":1417,"originalUrl":1418,"categories":1419},"\u002Fblog\u002Ffield-report-the-reality-of-building-agentic-ai-in-2025","Field Report. The Reality of Building Agentic AI in 2025","From Hype to Sovereign Infrastructure Nederlandse Versie Summary The narrative surrounding “Agentic AI” in 2025 is defined by a sharp contrast between market expectations and engineering reality. While the general public, conditioned by the ease of ChatGPT, expects “miracles” and instant integration, the reality of building autonomous agents for enterprise workflows is a discipline of ...","2025-11-25T14:03:39","field-report-the-reality-of-building-agentic-ai-in-2025","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ffieldreport.webp","https:\u002F\u002Feliovp.com\u002Ffield-report-the-reality-of-building-agentic-ai-in-2025\u002F",[1321,1297,1420,1379,1421,1422,1423,1424,1425,1426,1427,1428,1352,1429],"Solutions","Agentic AI","AI Engineering","AI Strategy","Autonomous Agents","Enterprise AI","Local LLM","Model Fine-Tuning","On-Premise AI","VRAM Optimization",{"path":1431,"title":1432,"description":1433,"date":1434,"slug":1435,"image":1436,"originalUrl":1437,"categories":1438},"\u002Fblog\u002Fthe-synthetic-unicorn-bubble","The Synthetic Unicorn Bubble","How AI Neocloud Infrastructure Turns Circular Cash Flows Into Fake Growth Executive Summary The interval between 2023 and 2025 has birthed a capital allocation phenomenon arguably without precedent: the “Synthetic Bubble.” Driven by the scramble for AI dominance, the venture capital apparatus has directed billions into the “Neocloud” ecosystem. However, a rigorous analysis suggests that ...","2025-11-24T19:22:28","the-synthetic-unicorn-bubble","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fsyntheticunicorn.webp","https:\u002F\u002Feliovp.com\u002Fthe-synthetic-unicorn-bubble\u002F",[1321,1297,1379,1322,1439,1440,1441,1442,1443,1444,1445,1446,1447],"AI Neocloud","Circular Financing","GPU Cloud","Investment Risks","Startup Valuation","Synthetic Bubble","Tech Analysis","Vaporware","Venture Capital",{"path":1449,"title":1450,"description":1451,"date":1452,"slug":1453,"image":1454,"originalUrl":1455,"categories":1456},"\u002Fblog\u002Fbuilding-the-engine-for-the-ai-race-the-4-month-path-to-nvidia-gb300-nvl72-power","Building the Engine for the AI Race: The 4-Month Path to NVIDIA GB300 NVL72 Power","In artificial intelligence infrastructure, speed is the foundation of competitive differentiation. From model training velocity to inference latency, every millisecond matters. But before any workload executes, there is a critical prerequisite that often determines success or failure: time to market. Traditional builds typically require 18–24 months. In the current AI cycle, that is simply too ...","2025-11-20T14:10:19","building-the-engine-for-the-ai-race-the-4-month-path-to-nvidia-gb300-nvl72-power","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fsuperpodmodflexfrontimage.webp","https:\u002F\u002Feliovp.com\u002Fbuilding-the-engine-for-the-ai-race-the-4-month-path-to-nvidia-gb300-nvl72-power\u002F",[1321,1367,1368,1457,1322,1458,1459,1460,1461,1462,1463,1464,1465],"150kW Rack","DLC","High Density","Liquid Cooling","Modular Data Center","NVIDIA Blackwell Ultra","NVIDIA GB300","NVL72","Rapid Deployment",{"path":1467,"title":1468,"description":1469,"date":1470,"slug":1471,"image":1472,"originalUrl":1473,"categories":1474},"\u002Fblog\u002Fwhy-cuda-translation-wont-unlock-amds-real-potential","Why “CUDA” Translation Won’t Unlock AMD’s Real Potential","Every few years, a new solution pops up promising the same dream: On paper, that sounds perfect. Take your existing CUDA applications, swap out the toolchain, and suddenly you’re “portable.” And to be fair: if you’re running research code or trying to get an internal tool to compile on a non-NVIDIA box, that can absolutely ...","2025-11-12T14:48:37","why-cuda-translation-wont-unlock-amds-real-potential","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fchatgpt-image-nov-11-2025-09_16_10-pm-1.webp","https:\u002F\u002Feliovp.com\u002Fwhy-cuda-translation-wont-unlock-amds-real-potential\u002F",[1321,1297,875,1368,1475,1297,1476,1477,1478,1479,1480,1481,875,1482],"AMD MI300X","CUDA Translation","FP8","GPU Optimization","High Performance Computing","HIP","Kernel Tuning","ROCm",{"path":1484,"title":1485,"description":1486,"date":1487,"slug":1488,"image":1489,"originalUrl":1490,"categories":1491},"\u002Fblog\u002Fpaiton-the-simplest-way-to-supercharge-ai-inference","Paiton: The Simplest Way to Supercharge AI Inference","Let’s be honest, we’re not the marketing type.We’ve never taken a cent of outside investment, never burned cash on ad campaigns, and never hired a sales army.We just build things that work. In today’s world, it seems the companies shouting the loudest often get the spotlight, while the ones doing the actual engineering quietly build ...","2025-11-11T10:31:22","paiton-the-simplest-way-to-supercharge-ai-inference","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaiton-powaaah.webp","https:\u002F\u002Feliovp.com\u002Fpaiton-the-simplest-way-to-supercharge-ai-inference\u002F",[1321,1297,875,1298,1492,1475,1303,1493,1301,1481,875,1494,1250],"AMD Instinct","High Throughput","SGLang",{"path":1496,"title":1497,"description":1498,"date":1499,"slug":1500,"image":1501,"originalUrl":1502,"categories":1503},"\u002Fblog\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens","Stop Overpaying: Paiton MI300X MoE Beats H200\u002FB200 on $\u002F1M Tokens","Short summary: We benchmarked Paiton with our new MoE support on Qwen\u002FQwen3-30B-A3B-Instruct-2507 to compare inference performance across several setups. Each configuration was run five times per batch size and we report the mean across runs. Why this benchmark Most published numbers use synthetic prompts or toy datasets. We focused on realistic conversational workloads (we always ...","2025-09-26T13:36:18","stop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fhulkvshulkpaitonwins.webp","https:\u002F\u002Feliovp.com\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens\u002F",[1321,1297,875,1504,1475,1505,1301,1506,1507,1508,1509,875,1510],"AI Benchmarks","Cost per Token","Mixture of Experts","MoE","NVIDIA B200","NVIDIA H200","Qwen3",{"path":1512,"title":1513,"description":1514,"date":1515,"slug":1516,"image":1517,"originalUrl":1518,"categories":1519},"\u002Fblog\u002Fagentic-ai-but-make-it-local-from-inbox-to-insight-to-action-en","Agentic AI, But Make It Local: From Inbox to Insight to Action","(Nederlandse versie) We’ve built production-ready, local-first agentic AI that plugs into your existing email stack, auto-creates tickets, classifies messages, extracts multi-question threads, reads PDFs, spots invoices\u002Fquotes, analyzes images (yes, damage detection), and pushes structured reports into your systems, no dependency on OpenAI, Google, or Microsoft unless you want it. Tailor-made models trained on your data, ...","2025-09-16T13:09:00","agentic-ai-but-make-it-local-from-inbox-to-insight-to-action-en","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ffrontfotoblog.webp","https:\u002F\u002Feliovp.com\u002Fagentic-ai-but-make-it-local-from-inbox-to-insight-to-action-en\u002F",[1321,1297,1420,1368,1421,1520,1521,1522,1523,1426,1428,1352,1524,1525],"Damage Detection","Document Processing","Email Automation","Invoice Extraction","Ticket Automation","Workflow Automation",{"path":1527,"title":1528,"description":1529,"date":1530,"slug":1531,"image":1532,"originalUrl":1533,"categories":1534},"\u002Fblog\u002Fmi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach","MI300X FP8 Data‑Parallel Benchmarks (8–64 GPUs): H200 Left Behind, B200 Within Reach","At ElioVP, we’re all about pushing AI inference past the limits, and packaging every squeeze of performance into a plug‑and‑play runtime.  Remember our last blog, where Paiton’s FP8 pipeline on AMD’s MI300X completely outclassed NVIDIA’s H200? Well, buckle up, because we’ve gone back to the drawing board. This time, we’re loading Llama-3.1-8B-Instruct-FP8-KV, the leaner, meaner ...","2025-07-31T13:32:57","mi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fimage-2-1.webp","https:\u002F\u002Feliovp.com\u002Fmi300x-fp8-data%e2%80%91parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach\u002F",[1321,1297,875,1368,1535,1337,1338,1536,1537,1348,1349,875,1250],"AI","H200","MI300X",{"path":1539,"title":1540,"description":1541,"date":1542,"slug":1543,"image":1544,"originalUrl":1545,"categories":1546},"\u002Fblog\u002Fapplicable-ai-for-businesses","Applicable AI for Businesses","Here at Eliovp, we continuously innovate when it comes to building practical solutions. If there’s one core strength, it’s our team’s ability to think outside the box. One key area of focus for us is developing applicable AI solutions, everyday usable AI implementations tailored specifically for our clients’ needs. In this blog, we’ll explore our ...","2025-07-09T21:35:30","applicable-ai-for-businesses","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fscherm_afbeelding-2025-07-09-om-23.30.17.webp","https:\u002F\u002Feliovp.com\u002Fapplicable-ai-for-businesses\u002F",[1321,1297,1420,1421,1520,1521,1522,1523,1426,1428,1352,1524,1525],{"path":1548,"title":1549,"description":1550,"date":1551,"slug":1552,"image":25,"originalUrl":1553,"categories":1554},"\u002Fblog\u002Fintroducing-paitons-free-evaluation-models","Introducing Paiton’s Free Evaluation Models","Introduction AI is rapidly transforming every industry, but running large models efficiently remains a major technical and financial challenge. At ElioVP, we specialize in optimizing for AMD accelerators, helping organizations unlock the full potential of their hardware. Today, we’re excited to announce a new offering: free evaluation models that let you test our cutting-edge optimizations ...","2025-07-07T11:26:13","introducing-paitons-free-evaluation-models","https:\u002F\u002Feliovp.com\u002Fintroducing-paitons-free-evaluation-models\u002F",[1321,1297,875],{"path":1556,"title":1557,"description":1558,"date":1559,"slug":1560,"image":1561,"originalUrl":1562,"categories":1563},"\u002Fblog\u002Fpaiton-dramatically-faster-startup-and-performance-for-llama-3-1-405b","Paiton: Dramatically Faster Startup and Performance for Llama-3.1-405B","With Paiton, we’re not merely pursuing peak inference speeds, we’re fundamentally reshaping the entire lifecycle of large language model (LLM) deployment. Our latest endeavor pairs AMD’s cutting-edge MI300X GPUs with the colossal Llama-3.1-405B-Instruct-FP8-KV model, achieving groundbreaking milestones: Visual Demonstration: Startup Speed Showcase We’re excited to share a visual demonstration of Paiton’s revolutionary startup performance. Watch ...","2025-06-12T20:15:23","paiton-dramatically-faster-startup-and-performance-for-llama-3-1-405b","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fservingscreenshot.webp","https:\u002F\u002Feliovp.com\u002Fpaiton-dramatically-faster-startup-and-performance-for-llama-3-1-405b\u002F",[1321,1297,875,1368,1298,1475,1564,1477,1565,1566,1567,875,1568,1569],"Cold Start","Graph Compilation","Llama 3.1 405B","LLM Optimization","Startup Latency","Tensor Parallelism",{"path":1571,"title":1572,"description":1573,"date":1574,"slug":1575,"image":1576,"originalUrl":1577,"categories":1578},"\u002Fblog\u002Fpaiton-fp8-beats-nvidias-h200-on-amds-mi300x","Paiton FP8 Beats NVIDIA’s H200 on AMD’s MI300X","The world of AI is moving at an unprecedented pace, and efficient inference is key to deploying powerful models in real-world applications. At Eliovp, we’ve consistently pushed the boundaries of AI performance, as highlighted in our previous blogs showcasing significant inference speedups when benchmarking with fp16\u002Fbf16. Now, we’re thrilled to announce a further significant leap ...","2025-06-08T19:12:40","paiton-fp8-beats-nvidias-h200-on-amds-mi300x","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fblognewfp8.webp","https:\u002F\u002Feliovp.com\u002Fpaiton-fp8-beats-nvidias-h200-on-amds-mi300x\u002F",[1321,1297,875,1368,1475,1579,1425,1344,1299,1300,1302,1566,1580,1581],"Cold Start Optimization","Model Serving","vLLM Optimization",{"path":1583,"title":1584,"description":1585,"date":1586,"slug":1587,"image":1588,"originalUrl":1589,"categories":1590},"\u002Fblog\u002Fmi300x-vs-h200-vs-rx-7900-xtx-vs-tenstorrent-n300s-with-vllm","MI300X vs H200 vs RX 7900 XTX vs Tenstorrent n300s with vLLM","As large language models (LLMs) become a foundational part of modern applications, picking the right server for deployment is more important than ever. Whether you’re an enterprise scaling up inference, a startup optimizing for cost, or a researcher pushing throughput boundaries. This blog compares two high-profile server setups and two not so high-profile setups which ...","2025-05-09T14:03:58","mi300x-vs-h200-vs-rx-7900-xtx-vs-tenstorrent-n300s-with-vllm","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fcomparisontenstor.webp","https:\u002F\u002Feliovp.com\u002Fmi300x-vs-h200-vs-rx-7900-xtx-vs-tenstorrent-n300s-with-vllm\u002F",[1321,1297,875,1420,1368,1337,1537,1349,1591,1592],"RX7900XTX","tenstorrent",{"path":1594,"title":1595,"description":1596,"date":1597,"slug":1598,"image":1599,"originalUrl":1600,"categories":1601},"\u002Fblog\u002Fclusterpl-empowering-gpu-cluster-investors-with-real-world-financial-insights","ClusterP&L: Empowering GPU Cluster Investors with Real-World Financial Insights","At Eliovp BV, we’ve spent years on the cutting edge of GPU cluster deployment and optimization across Europe. Our team supports leading organizations in AI, finance, and research, architecting, building, and scaling high-performance infrastructure. Over time, our customers, both newcomers and seasoned adopters, repeatedly asked the same question: “Can you help us build a P&L ...","2025-05-03T10:52:22","clusterpl-empowering-gpu-cluster-investors-with-real-world-financial-insights","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fcomparisonscenarios.webp","https:\u002F\u002Feliovp.com\u002Fclusterpl-empowering-gpu-cluster-investors-with-real-world-financial-insights\u002F",[1321,1297,1367,1420,1338,1536,1602,1349,1603],"MI325x","pnl calculator",{"path":1605,"title":1606,"description":1607,"date":1608,"slug":1609,"image":1610,"originalUrl":1611,"categories":1612},"\u002Fblog\u002Fcranking-out-faster-tokens-for-fewer-dollars-amd-mi300x-vs-nvidia-h200","Cranking Out Faster Tokens for Fewer Dollars: AMD MI300X vs. NVIDIA H200","Qwen3-32B on Paiton + AMD MI300x vs.NVIDIA H200 1. Introduction “While we’re actively training models for local customers, automating and streamlining critical business processes, we still found time to push our Paiton framework to the limit on Qwen3-32B.” In the competitive realm of LLMs, next-gen hardware like the NVIDIA H200 often steals the headlines. But ...","2025-05-02T21:10:30","cranking-out-faster-tokens-for-fewer-dollars-amd-mi300x-vs-nvidia-h200","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002F3ac59a73-2466-4422-b7e5-ef2e4a8ca58e.webp","https:\u002F\u002Feliovp.com\u002Fcranking-out-faster-tokens-for-fewer-dollars-amd-mi300x-vs-nvidia-h200\u002F",[1321,1297,875,1535,1337,1536,1613,1349,875,1250],"MI300",{"path":1615,"title":1616,"description":1617,"date":1618,"slug":1619,"image":1620,"originalUrl":1621,"categories":1622},"\u002Fblog\u002Fpower-meets-precision-high-density-modular-data-center-for-nvidia-nvl-deployments-1-2-mw","Power Meets Precision: High-Density Modular Data Center for NVIDIA NVL Deployments (1–2 MW)","Purpose-Built High-Density Infrastructure for Blackwell-Class AI Workloads At Eliovp, we’re engineering a new class of AI infrastructure. Our advanced modular platform is built to also support NVIDIA’s cutting-edge NVL architecture, from the efficient NVL4 to the ultra-scale NVL72, enabling deployments that range from distributed edge inference to full-stack model training at hyperscale. Designed to meet ...","2025-05-02T14:09:59","power-meets-precision-high-density-modular-data-center-for-nvidia-nvl-deployments-1-2-mw","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Feliovp_critical-1mw-pod_rev-2_transparent.webp","https:\u002F\u002Feliovp.com\u002Fpower-meets-precision-high-density-modular-data-center-for-nvidia-nvl-deployments-1-2-mw\u002F",[1321,1367,1623,1322,1459,1325,1460,1461,1624,1625,1464,1626],"1-2MW Data Center","NVIDIA Blackwell","NVIDIA NVL","Precision Cooling",{"path":1628,"title":1629,"description":1630,"date":1631,"slug":1632,"image":1633,"originalUrl":1634,"categories":1635},"\u002Fblog\u002Fexamining-ai-agents-in-the-medical-field-ai-that-speaks-dicom","Examining AI agents in the medical field: AI that speaks DICOM","At Eliovp, we’re constantly keeping up with the newest AI trends. Consequently, we have been looking into AI agents and have created a medical agent designed to seamlessly interact with DICOM servers inside hospitals. This isn’t just another chatbot or AI tool. This is an intelligent assistant that understands the language of radiology and is ...","2025-04-11T14:45:48","examining-ai-agents-in-the-medical-field-ai-that-speaks-dicom","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fhealthcareblog-1.webp","https:\u002F\u002Feliovp.com\u002Fexamining-ai-agents-in-the-medical-field-ai-that-speaks-dicom\u002F",[1321,1297,1420,1368,1535,1337,1636],"Healthcare",{"path":1638,"title":1639,"description":1640,"date":1641,"slug":1642,"image":1643,"originalUrl":1644,"categories":1645},"\u002Fblog\u002Feliovp-bv-your-trusted-partner-for-supply-chain-resilience-amidst-new-u-s-tariffs","Eliovp BV: Your Trusted Partner for Supply Chain Resilience Amidst New U.S. Tariffs","In today’s rapidly evolving global trade landscape, businesses face unprecedented challenges in maintaining efficient and cost-effective IT infrastructure. The recent U.S. tariff adjustments have created waves of uncertainty across international markets, particularly for companies relying on high-performance computing and AI solutions. At Eliovp BV, we want to assure our valued clients that our comprehensive end-to-end ...","2025-04-04T10:01:27","eliovp-bv-your-trusted-partner-for-supply-chain-resilience-amidst-new-u-s-tariffs","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ftariffsshipping.webp","https:\u002F\u002Feliovp.com\u002Feliovp-bv-your-trusted-partner-for-supply-chain-resilience-amidst-new-u-s-tariffs\u002F",[1321,1379,1535,1337,1646,1647,1648,1649],"import","Taiwan","Tariffs","Trump",{"path":1651,"title":1652,"description":1653,"date":1654,"slug":1655,"image":1656,"originalUrl":1657,"categories":1658},"\u002Fblog\u002Fwhy-ai-agents-are-the-future","Why AI Agents Are the Future","1. Versatile IntegrationAI Agents are designed to integrate seamlessly with your existing software stack. This includes ERP, CRM, and marketing automation platforms. Instead of disrupting current systems, they complement and enhance them, all while learning from and adapting to your specific operational needs. 2. Intelligent Decision-MakingConventional automation scripts handle if-then scenarios, but they fall short ...","2025-03-23T22:06:59","why-ai-agents-are-the-future","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ferp2.jpeg","https:\u002F\u002Feliovp.com\u002Fwhy-ai-agents-are-the-future\u002F",[1321,1297,1420,1535,1659,1660],"AI Agents","ERP",{"path":1662,"title":1663,"description":1664,"date":1665,"slug":1666,"image":1667,"originalUrl":1668,"categories":1669},"\u002Fblog\u002Fthe-rise-of-open-source-ai-model-optimization","The Rise of Open-Source AI Model Optimization","In the rapidly evolving landscape of artificial intelligence (AI), open-source solutions are emerging as pivotal drivers of innovation and performance enhancement. These community-driven platforms democratize access to cutting-edge technologies, fostering collaboration and accelerating advancements in AI model optimization. The Open-Source Revolution in AI Open-source AI models have transformed the development and deployment of machine learning ...","2025-03-22T20:59:23","the-rise-of-open-source-ai-model-optimization","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Friseofopensource.jpeg","https:\u002F\u002Feliovp.com\u002Fthe-rise-of-open-source-ai-model-optimization\u002F",[1321,1297,1379,1670,1337,546,1349],"AI news",{"path":1672,"title":1673,"description":1674,"date":1675,"slug":1676,"image":1677,"originalUrl":1678,"categories":1679},"\u002Fblog\u002Fintroducing-our-benchmarking-tool-powered-by-dstack","Introducing Our Benchmarking Tool: Powered by dstack","1. Introduction Benchmarking is an essential part of optimizing AI models and software applications. Whether you’re testing AI model inference speeds, profiling different hardware configurations, or ensuring system performance over time, having a reliable benchmarking tool is crucial. However, many existing tools suffer from issues like inconsistent environments, difficult configuration setups, and lack of automation. ...","2025-03-20T14:21:59","introducing-our-benchmarking-tool-powered-by-dstack","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fbenchmarktool.jpeg","https:\u002F\u002Feliovp.com\u002Fintroducing-our-benchmarking-tool-powered-by-dstack\u002F",[1321,1297,875,1535,1337,1680,1681,1537,875],"benchmark","LLM",{"path":1683,"title":1684,"description":1685,"date":1686,"slug":1687,"image":1688,"originalUrl":1689,"categories":1690},"\u002Fblog\u002Foptimizing-qwq-32b-by-qwen-amd-mi300x-vs-nvidia-h200","Optimizing QwQ-32B (by Qwen): AMD MI300X vs. NVIDIA H200","1. Introduction In the world of large language models (LLMs), most benchmarks center on Llama or DeepSeek derivatives. We decided to diversify by adding the Qwen2 architecture, using our Paiton framework. This 32-billion-parameter model pushes GPU resources to the limit, perfect for comparing NVIDIA’s new H200 to our AMD MI300X, which leverages Paiton for advanced ...","2025-03-19T21:41:44","optimizing-qwq-32b-by-qwen-amd-mi300x-vs-nvidia-h200","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaiton4.jpeg","https:\u002F\u002Feliovp.com\u002Foptimizing-qwq-32b-by-qwen-amd-mi300x-vs-nvidia-h200\u002F",[1321,1297,875],{"path":1692,"title":1693,"description":1694,"date":1695,"slug":1696,"image":1697,"originalUrl":1698,"categories":1699},"\u002Fblog\u002Feliovp-featured-on-amd-tech-talk-podcast","Eliovp Featured on AMD “Tech Talk” Podcast","We’re excited to share that Eliovp was recently featured on AMD’s “Tech Talk” podcast! In this episode, our CEO, Elio Van Puyvelde sits down with Jim greene to talk about the origins of Eliovp, the passion and expertise that brought the company to life, and the innovative full end-to-end solutions we offer today. From our ...","2025-03-19T07:53:39","eliovp-featured-on-amd-tech-talk-podcast","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ftechtalkjimgreene.jpeg","https:\u002F\u002Feliovp.com\u002Feliovp-featured-on-amd-tech-talk-podcast\u002F",[1321,1337,1700,1701,1702],"Jim Greene","Podcast","Tech Talk",{"path":1704,"title":1705,"description":1706,"date":1707,"slug":1708,"image":1709,"originalUrl":1710,"categories":1711},"\u002Fblog\u002Ffurther-optimizing-amd-powered-inference-with-paiton","Further Optimizing AMD-Powered Inference with Paiton","Executive Summary If you’ve followed our journey so far, you’ll know that Paiton is laser-focused on AMD-centric inference optimization. Our latest work takes DeepSeek R1 Distill Llama 8B to the next level, delivering 10–15% higher throughput, improved time-to-first-token (TTFT), and more stable performance at lower batch sizes, an area that previously needed a boost. In short, Paiton further cements its ability to exploit AMD hardware’s raw power, bridging ...","2025-03-13T06:18:30","further-optimizing-amd-powered-inference-with-paiton","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaitonpost3.webp","https:\u002F\u002Feliovp.com\u002Ffurther-optimizing-amd-powered-inference-with-paiton\u002F",[1321,1297,875,1337,1712,1713,1536,1537,1602,875,1250],"Deepseek","H100",{"path":1715,"title":1716,"description":1717,"date":1718,"slug":1719,"image":1720,"originalUrl":1721,"categories":1722},"\u002Fblog\u002Fa-first-look-at-paiton-in-action-deepseek-r1-distill-llama-3-1-8b","A First Look at Paiton in Action: Deepseek R1 Distill Llama 3.1 8B","Outperforming Stock Models on the AMD MI300X 1. Introduction We couldn’t wait to show what Paiton can really do. After detailing our AMD-centric approach and architecture-level optimizations in our previous blog post, we decided to test-drive Paiton on a hype-worthy model: Deepseek R1 Distill Llama 3.1 8B. By compiling the model into efficient libraries and fusing ...","2025-01-31T09:11:02","a-first-look-at-paiton-in-action-deepseek-r1-distill-llama-3-1-8b","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaitonpost2.webp","https:\u002F\u002Feliovp.com\u002Fa-first-look-at-paiton-in-action-deepseek-r1-distill-llama-3-1-8b\u002F",[1321,1297,875,1337,1712,1713,1536,1537,1602,875,1250],{"path":1724,"title":1725,"description":1726,"date":1727,"slug":1728,"image":1729,"originalUrl":1730,"categories":1731},"\u002Fblog\u002Fai-model-optimization-with-paiton","AI Model Optimization with Paiton","In the fast-paced world of artificial intelligence, model efficiency and performance are paramount. At ElioVP, we’re redefining what’s possible by delivering unparalleled optimization solutions for AI models with Paiton. By compiling the model’s architecture and leveraging our custom-written kernels, Paiton enables faster inference and reduced resource consumption on AMD GPUs. Why Model Optimization is More ...","2025-01-30T19:53:25","ai-model-optimization-with-paiton","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaitonpost1.webp","https:\u002F\u002Feliovp.com\u002Fai-model-optimization-with-paiton\u002F",[1321,1297,875,1337,1713,1536,1537,1602,875,1250],1789394884406]