MORE FROM AMD GPUS. PROVEN ON REAL MODELS.
Paiton finds what slows your model down, replaces the generic runtime path,
and turns AMD hardware into faster production inference without retraining.
MEASURED ON REAL GENERATION AND INFERENCE PATHS.
The gap is rarely the model. It is usually the runtime.
Paiton tunes the exact path your workload uses, then proves the gain with benchmarks.
Local image generation: Radeon AI PRO R9700
FLUX.2 klein 4B generates 1024 × 1024 images in four steps on one R9700, without CPU offload. Warm prompt-to-PIL timing excludes startup, PNG saving and the interface. Free ComfyUI workflow included.
Read the FLUX.2 klein benchmarkVideo diffusion: MI355X vs B200
Paiton-Diffusers removed runtime waste from the generation path and let the hardware show up.
Read Wan2.2 BenchmarkYou do not need a new model to improve unit economics. You need the runtime, kernels, and deployment path tuned around the workload you already run.
TRANSLATE THROUGHPUT INTO REUSABLE CAPACITY.
EDITABLE PLANNING SCENARIO
Faster inference is more than a benchmark number. It can reduce the GPU time required for the same workload,
or increase the output of hardware you already own.
Start with this illustrative scenario, then replace the assumptions with the uplift measured during your Paiton benchmark.
LLMS, VIDEO DIFFUSION, MOE, AND CUSTOM STACKS.
Whether you run LLMs, image and video diffusion, MoE, or custom architectures,
Paiton starts where money is lost: latency, throughput, memory pressure, and cost per token.
More tokens, less waiting
Find the bottleneck that users and invoices feel. Qwen3.8 on one Radeon AI PRO R9700 delivered 21% more interactive output and 54.3% more coding output.
No retraining required
Keep the model and improve the path around it with AMD-aware operators and custom .so files.
Runtime work that pays back
Local FLUX.2 klein image generation with ComfyUI on R9700, Wan2.2 video generation on Instinct, and vLLM support for language inference.
Language Models
Llama, Qwen, Deepseek, Gemma, Mistral, and CodeLlama with vLLM support, FP8 precision, and custom kernels when the gain is worth it.
Diffusion & Video
Wan2.2, Flux, Stable Diffusion, SDXL, ControlNet, and text-to-video workloads where scheduler, memory, and kernel choices move the number.
Advanced Models
MoE, MLA, multimodal systems, proprietary architectures, expert kernels, novel attention, and custom inference stacks.
MEASURE. TUNE. PROVE.
Bring the workload, the target GPU, and the performance goal.
We turn that into a focused path from measurement to production-ready gain.
Run the real workload and identify where latency, memory pressure, or cost per token is leaking value.
Build the AMD-specific path with custom .so files, FP8 precision, kernel fusion, and the same model weights.
Deploy the optimized path on AMD Instinct or Radeon AI PRO hardware and compare the result against the baseline.
FP8 can reduce memory pressure while preserving accuracy. Tensor parallelism spreads large models like Llama-3.1-405B across multiple AMD GPUs. Custom kernels target the bottlenecks generic runtimes leave behind.
MI355X
288GB HBM3E
MI325X / MI300X / MI300A
256GB HBM3E, 192GB HBM3, 128GB HBM3 APU
MI250X / MI250 / MI210
Datacenter AMD GPU support
MI100
First compute-optimized generation
Radeon AI PRO R9700
Qualified Qwen3.8, Ornith 1.5 and FLUX.2 klein profiles on 32GB GDDR6
RX 7900 XTX / RX 7900 XT
Supported for selected inference workloads
RX 6800 XT / RX 6900 XT
Supported with workload-specific qualification
Paiton supports both AMD Instinct datacenter accelerators and Radeon GPUs. Every model, runtime, precision, and hardware combination is qualified against a defined deployment scope before delivery.
OUR CONTRIBUTION TO OPEN SOURCE.
Better inference should reach beyond our customer projects. Our open-source Paiton vLLM plugin brings model-specific AMD GPU optimizations to developers running AI on their own hardware.
Explore the code, run the supported models locally, and share what you learn. It is our way of helping more people get more from the GPUs they already own.
Code, containers and setup guides. Ready to explore on GitHub.
Explore the plugin on GitHubPublished profiles cover Qwen3.8 and Ornith 1.5 on Radeon AI PRO R9700, for single-user text inference. See the repository for exact configurations and requirements.
16.2% less generation time
1.054 seconds per image with a warm pipeline, 33.4% lower peak Torch allocation, and a free local ComfyUI workflow. Tested at 1024 × 1024, four steps and batch one.
Read Benchmark+27% output
Paiton with DFlash delivers 44.63 output tokens per second on R9700, 27% ahead of tuned stock vLLM in the tested single-user workload.
Read Benchmark+21% output
Higher interactive, coding, and long-context throughput from the same 32GB Radeon GPU.
Read Benchmark17.6% faster
Text-to-video diffusion speedup with Paiton-Diffusers.
Read Benchmark405B model
Faster startup and inference for a frontier-scale open model.
Read Case Study$/1M tokens
MI300X plus Paiton runtime on long-context MoE economics.
Read Benchmark
Share the model, target GPU, and production goal.
Paiton turns that into a benchmark-led optimization plan.
