AI chips and platforms: specs, runtimes, measured speeds
Every AI chip and platform we track: memory, key specs, which runtimes run on it with evidence, and the configs we measured.
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.
80 chips and platforms.
- Cerebras WSE-3 (Cerebras Inference) · Cerebras · cloud-api · 0 measured configs
- Groq LPU (GroqCloud) · Groq · cloud-api · 0 measured configs
- SambaNova SN40L (SambaCloud) · SambaNova · cloud-api · 0 measured configs
- AMD Radeon RX 7900 XTX · AMD · consumer-gpu · 0 measured configs
- AMD Radeon RX 9070 XT · AMD · consumer-gpu · 0 measured configs
- Intel Arc B580 (12GB) · Intel · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 3060 12GB · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 3090 · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 4060 Ti (16GB) · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 4070 Ti SUPER · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 4080 · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 4080 SUPER · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 5060 Ti (16GB) · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 5070 Ti · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA GeForce RTX 5080 · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA RTX 4090 (24GB) · NVIDIA · consumer-gpu · 0 measured configs
- NVIDIA RTX 5090 (32GB) · NVIDIA · consumer-gpu · 207 measured configs
- AWS Inferentia2 · AWS · datacenter-accelerator · 0 measured configs
- AWS Trainium2 · AWS · datacenter-accelerator · 0 measured configs
- AWS Trainium3 · AWS · datacenter-accelerator · 0 measured configs
- FuriosaAI RNGD · FuriosaAI · datacenter-accelerator · 0 measured configs
- Google Cloud TPU v6e · Google · datacenter-accelerator · 0 measured configs
- Google Cloud TPU7x (Ironwood) · Google · datacenter-accelerator · 0 measured configs
- Huawei Ascend 310P · Huawei · datacenter-accelerator · 0 measured configs
- Huawei Ascend 910C · Huawei · datacenter-accelerator · 0 measured configs
- Intel Gaudi 3 · Intel · datacenter-accelerator · 0 measured configs
- PFN MN-Core 2 · Preferred Networks · datacenter-accelerator · 0 measured configs
- Rebellions ATOM · Rebellions · datacenter-accelerator · 0 measured configs
- AMD Instinct MI300X (192GB) · AMD · datacenter-gpu · 0 measured configs
- AMD Instinct MI325X · AMD · datacenter-gpu · 0 measured configs
- AMD Instinct MI355X · AMD · datacenter-gpu · 0 measured configs
- NVIDIA A10 · NVIDIA · datacenter-gpu · 12 measured configs
- NVIDIA A100 40GB · NVIDIA · datacenter-gpu · 61 measured configs
- NVIDIA A100 80GB · NVIDIA · datacenter-gpu · 0 measured configs
- NVIDIA B200 · NVIDIA · datacenter-gpu · 0 measured configs
- NVIDIA B300 (Blackwell Ultra) · NVIDIA · datacenter-gpu · 0 measured configs
- NVIDIA H100 PCIe · NVIDIA · datacenter-gpu · 73 measured configs
- NVIDIA H100 SXM · NVIDIA · datacenter-gpu · 70 measured configs
- NVIDIA H200 · NVIDIA · datacenter-gpu · 0 measured configs
- NVIDIA L40S · NVIDIA · datacenter-gpu · 0 measured configs
- Axelera Europa · Axelera AI · edge · 0 measured configs
- Axelera Metis · Axelera AI · edge · 0 measured configs
- EdgeCortix SAKURA-II · EdgeCortix · edge · 0 measured configs
- Google Coral Edge TPU · Google · edge · 0 measured configs
- Hailo-10H · Hailo · edge · 0 measured configs
- Hailo-8 · Hailo · edge · 0 measured configs
- Hailo-8L · Hailo · edge · 0 measured configs
- MediaTek Genio 720 · MediaTek · edge · 0 measured configs
- NVIDIA Jetson AGX Orin 64GB · NVIDIA · edge · 0 measured configs
- NVIDIA Jetson AGX Thor · NVIDIA · edge · 0 measured configs
- NVIDIA Jetson Orin Nano Super (8GB) · NVIDIA · edge · 0 measured configs
- NVIDIA Jetson Orin NX 16GB · NVIDIA · edge · 0 measured configs
- Rockchip RK3576 · Rockchip · edge · 0 measured configs
- Rockchip RK3588 · Rockchip · edge · 0 measured configs
- ST STM32N6 · STMicroelectronics · edge · 0 measured configs
- Qualcomm Cloud AI 100 Ultra · Qualcomm · npu-accelerator · 0 measured configs
- Tenstorrent Blackhole p150 · Tenstorrent · npu-accelerator · 0 measured configs
- Tenstorrent Wormhole n150 · Tenstorrent · npu-accelerator · 0 measured configs
- Tenstorrent Wormhole n300 · Tenstorrent · npu-accelerator · 0 measured configs
- Intel Core Ultra 200V (Lunar Lake) · Intel · pc-soc · 0 measured configs
- Intel Core Ultra Series 3 (Panther Lake) · Intel · pc-soc · 0 measured configs
- Qualcomm Snapdragon X2 Elite · Qualcomm · pc-soc · 0 measured configs
- AMD Gorgon Halo (Ryzen AI Max+ PRO 495) · AMD · unified-memory-box · 0 measured configs
- AMD Strix Halo (Ryzen AI Max+ 395) · AMD · unified-memory-box · 212 measured configs
- Apple M-series Max (MacBook Pro / Mac Studio Max) · Apple · unified-memory-box · 29 measured configs
- Apple M1 Max · Apple · unified-memory-box · 0 measured configs
- Apple M1 Ultra · Apple · unified-memory-box · 0 measured configs
- Apple M2 Max · Apple · unified-memory-box · 0 measured configs
- Apple M2 Ultra · Apple · unified-memory-box · 0 measured configs
- Apple M3 Max (40-core GPU) · Apple · unified-memory-box · 0 measured configs
- Apple M3 Ultra · Apple · unified-memory-box · 0 measured configs
- Apple M4 Max (40-core GPU) · Apple · unified-memory-box · 0 measured configs
- Apple M4 Pro · Apple · unified-memory-box · 0 measured configs
- Apple M5 Max (40-core GPU) · Apple · unified-memory-box · 0 measured configs
- Apple Mac Studio (M-series Ultra) · Apple · unified-memory-box · 17 measured configs
- NVIDIA DGX Spark (GB10) · NVIDIA · unified-memory-box · 164 measured configs
- Intel Arc Pro B60 (24GB) · Intel · workstation-gpu · 0 measured configs
- NVIDIA RTX 6000 Ada (48GB) · NVIDIA · workstation-gpu · 0 measured configs
- NVIDIA RTX A6000 · NVIDIA · workstation-gpu · 0 measured configs
- NVIDIA RTX PRO 6000 Blackwell (96GB) · NVIDIA · workstation-gpu · 0 measured configs
More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.