AI SDKs and runtimes: which chips each one runs on
The SDKs, runtimes and compilers that run models on chips, with the evidence for every chip each one supports.
Snapshot 2026-10-15 · 394 configs on AI boxes and 423 on GPU cards · 123 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.
50 SDKs and runtimes.
- ROCm · AMD · runs on 0 chips
- Ryzen AI Software · AMD · runs on 0 chips
- Core ML · Apple · runs on 0 chips
- MLX · Apple · runs on 0 chips
- MLX LM · Apple ML Explore (ml-explore/mlx-lm) · runs on 10 chips
- AWS Neuron SDK · AWS · runs on 0 chips
- Axelera Voyager SDK · Axelera AI · runs on 0 chips
- Cerebras Inference API · Cerebras · runs on 0 chips
- EdgeCortix MERA · EdgeCortix · runs on 0 chips
- Furiosa SDK / furiosa-llm · FuriosaAI · runs on 0 chips
- llama.cpp · ggml-org (ggml-org/llama.cpp) · runs on 41 chips
- JAX · Google · runs on 0 chips
- LiteRT (successor to TensorFlow Lite) · Google · runs on 0 chips
- vLLM TPU (tpu-inference) · Google / vLLM project · runs on 0 chips
- GroqCloud API · Groq · runs on 0 chips
- Hailo Dataflow Compiler · Hailo · runs on 0 chips
- HailoRT · Hailo · runs on 0 chips
- Huawei CANN · Huawei · runs on 0 chips
- Huawei MindIE · Huawei · runs on 0 chips
- Intel Gaudi Software · Intel · runs on 0 chips
- Intel oneAPI · Intel · runs on 0 chips
- OpenVINO · Intel · runs on 0 chips
- OpenVINO GenAI · Intel · runs on 4 chips
- SGLang · LMSYS / sgl-project (sgl-project/sglang) · runs on 18 chips
- MediaTek NeuroPilot · MediaTek · runs on 0 chips
- Metrale Engine · Metrale · runs on 1 chips
- ONNX Runtime · Microsoft (microsoft/onnxruntime) · runs on 59 chips
- CUDA Toolkit · NVIDIA · runs on 0 chips
- NVIDIA JetPack · NVIDIA · runs on 0 chips
- NVIDIA NIM for LLMs · NVIDIA · runs on 0 chips
- TensorRT · NVIDIA · runs on 0 chips
- TensorRT-LLM · NVIDIA · runs on 15 chips
- Ollama · Ollama (ollama/ollama) · runs on 33 chips
- halogen · peonist-ai · runs on 1 chips
- Qualcomm AI Hub · Qualcomm · runs on 0 chips
- Qualcomm AI Runtime (QAIRT / QNN) · Qualcomm · runs on 0 chips
- Qualcomm Genie · Qualcomm · runs on 0 chips
- Rebellions RBLN SDK · Rebellions · runs on 0 chips
- RKLLM (rknn-llm) · Rockchip · runs on 0 chips
- RKNN-Toolkit2 · Rockchip · runs on 0 chips
- SambaCloud API · SambaNova · runs on 0 chips
- ST Edge AI Core · STMicroelectronics · runs on 0 chips
- TT-Forge · Tenstorrent · runs on 0 chips
- tt-inference-server · Tenstorrent · runs on 0 chips
- TT-Metalium / TT-NN · Tenstorrent · runs on 0 chips
- vllm-tt-plugin · Tenstorrent · runs on 0 chips
- KTransformers · Tsinghua MADSys Lab / Approaching.AI (kvcache-ai/ktransformers) · runs on 13 chips
- ExLlamaV3 · turboderp (turboderp-org/exllamav3) · runs on 11 chips
- vLLM · vLLM project (vllm-project/vllm) · runs on 32 chips
- vLLM Ascend plugin · vLLM project / Huawei · runs on 0 chips
More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.