Can X run on Y? Measured answers

Only model and GPU pairs where someone actually measured the speed. No estimates. For anything else use the VRAM calculator.

Qwen3.8 27B

• Can 2x Apple M5 Ultra run Qwen3.8 27B? Yes: 33.27 tok/s
• Can 2x Intel Arc Pro B70 run Qwen3.8 27B? Yes: 46.01 tok/s
• Can 2x RTX 5070 Ti run Qwen3.8 27B? Yes: 230 tok/s
• Can 2x Tesla V100 32GB run Qwen3.8 27B? Yes: 104 tok/s
• Can 4x Apple M5 Ultra run Qwen3.8 27B? Yes: 50.89 tok/s
• Can Apple M3 Ultra run Qwen3.8 27B? Yes: 12.46 tok/s
• Can Apple M5 Ultra run Qwen3.8 27B? Yes: 19.69 tok/s
• Can DGX Spark (GB10) run Qwen3.8 27B? Yes: 19.0 tok/s
• Can Intel Arc Pro B70 run Qwen3.8 27B? Yes: 96 tok/s
• Can RTX 3060 12GB run Qwen3.8 27B? Partly: part of the model runs from system RAM (CPU offload, per the source): 5.2-6.7 tok/s
• Can RTX 3090 run Qwen3.8 27B? Yes: 139.7 tok/s
• Can RTX 4060 Ti 16GB run Qwen3.8 27B? Partly: part of the model runs from system RAM (CPU offload, per the source): 10.7-17.6 tok/s
• Can RTX 5070 Ti run Qwen3.8 27B? Yes: 139 tok/s
• Can RTX 5090 run Qwen3.8 27B? Yes: 75.2 tok/s
• Can RTX Pro 6000 Max-Q (Blackwell, 96 GB, 300 W) run Qwen3.8 27B? Yes: 150 tok/s
• Can RTX Spark N1X run Qwen3.8 27B? Yes: 14.5 tok/s
• Can Radeon AI Pro R9700 run Qwen3.8 27B? Yes: 57 tok/s
• Can Strix Halo (Radeon 8060S) run Qwen3.8 27B? Yes: 18.01 tok/s

Qwen3.6 35B-A3B

• Can 2x RTX 3090 run Qwen3.6 35B-A3B? Yes: 225-231 tok/s
• Can DGX Spark (GB10) run Qwen3.6 35B-A3B? Yes: 113 tok/s
• Can Intel Arc Pro B70 run Qwen3.6 35B-A3B? Yes: 120 tok/s
• Can RTX 3090 run Qwen3.6 35B-A3B? Yes: 146-153 tok/s
• Can RTX 4090 run Qwen3.6 35B-A3B? Yes: 160.6 tok/s
• Can RTX 5090 run Qwen3.6 35B-A3B? Yes: 258 tok/s
• Can RTX Spark N1X run Qwen3.6 35B-A3B? Yes: 89.5 tok/s
• Can Radeon AI Pro R9700 run Qwen3.6 35B-A3B? Yes: 117.69 tok/s
• Can Radeon RX 7900 XTX run Qwen3.6 35B-A3B? Yes: 135.2 tok/s
• Can Strix Halo (Radeon 8060S) run Qwen3.6 35B-A3B? Yes: 93 tok/s

Qwen3.6 27B

• Can DGX Spark (GB10) run Qwen3.6 27B? Yes: 32.83 tok/s
• Can RTX 3090 run Qwen3.6 27B? Yes: 42.7 tok/s
• Can RTX 4090 run Qwen3.6 27B? Yes: 47.0 tok/s
• Can RTX 5090 run Qwen3.6 27B? Yes: 75.7 tok/s
• Can Radeon AI Pro R9700 run Qwen3.6 27B? Yes: 66.0 tok/s
• Can Strix Halo (Radeon 8060S) run Qwen3.6 27B? Yes: 11.4 tok/s

Muse Spark Open 816B-A42B

• Can 2x Apple M5 Ultra run Muse Spark Open 816B-A42B? Yes: 24.99 tok/s
• Can 4x Apple M5 Ultra run Muse Spark Open 816B-A42B? Yes: 36.73 tok/s
• Can Apple M3 Ultra run Muse Spark Open 816B-A42B? Yes: 21 tok/s

Gemma 4 31B

• Can RTX 4090 run Gemma 4 31B? Yes: 43.8 tok/s
• Can RTX 5090 run Gemma 4 31B? Yes: 70.0 tok/s
• Can Strix Halo (Radeon 8060S) run Gemma 4 31B? Yes: 8.8 tok/s

Gemma 4 26B-A4B

• Can DGX Spark (GB10) run Gemma 4 26B-A4B? Yes: 56.5 tok/s
• Can Intel Arc Pro B70 run Gemma 4 26B-A4B? Yes: 122.16 tok/s
• Can RTX 4090 run Gemma 4 26B-A4B? Yes: 164 tok/s
• Can RTX 5090 run Gemma 4 26B-A4B? Yes: 235 tok/s
• Can Strix Halo (Radeon 8060S) run Gemma 4 26B-A4B? Yes: 51.8 tok/s

Gemma 4 12B

• Can RTX 3060 12GB run Gemma 4 12B? Yes: 38.5 tok/s
• Can RTX 4090 run Gemma 4 12B? Yes: 98.1 tok/s
• Can RTX 5070 Ti run Gemma 4 12B? Yes: 87.6 tok/s
• Can RTX 5090 run Gemma 4 12B? Yes: 142 tok/s

gpt-oss 120B

• Can Apple M2 Ultra run gpt-oss 120B? Yes: 79.68 tok/s
• Can DGX Spark (GB10) run gpt-oss 120B? Yes: 44.4 tok/s
• Can RTX Pro 6000 (Blackwell, 96 GB) run gpt-oss 120B? Yes: 196.31 tok/s
• Can Strix Halo (Radeon 8060S) run gpt-oss 120B? Yes: 44.7 tok/s

gpt-oss 20B

• Can Apple M1 Max run gpt-oss 20B? Yes: 75.15 tok/s
• Can Apple M1 Ultra run gpt-oss 20B? Yes: 100.43 tok/s
• Can Apple M2 Ultra run gpt-oss 20B? Yes: 116.08 tok/s
• Can Apple M3 Ultra run gpt-oss 20B? Yes: 115.52 tok/s
• Can Apple M4 Max run gpt-oss 20B? Yes: 92.36 tok/s
• Can DGX Spark (GB10) run gpt-oss 20B? Yes: 96.2 tok/s
• Can RTX 3060 12GB run gpt-oss 20B? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 83.96 tok/s
• Can RTX 3090 run gpt-oss 20B? Yes: 207.2 tok/s
• Can RTX 4070 Ti run gpt-oss 20B? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 163.14 tok/s
• Can RTX 4090 run gpt-oss 20B? Yes: 279.65 tok/s
• Can RTX 5060 Ti 16GB run gpt-oss 20B? Yes: 111.51 tok/s
• Can RTX 5070 Ti run gpt-oss 20B? Yes: 189.45 tok/s
• Can RTX 5080 run gpt-oss 20B? Yes: 204.85 tok/s
• Can RTX 5090 run gpt-oss 20B? Yes: 324.2 tok/s
• Can RTX A4000 run gpt-oss 20B? Yes: 93.55 tok/s
• Can RTX Pro 6000 (Blackwell, 96 GB) run gpt-oss 20B? Yes: 286.91 tok/s
• Can RTX Spark N1X run gpt-oss 20B? Yes: 58.0 tok/s
• Can Radeon AI Pro R9700 run gpt-oss 20B? Yes: 163.26 tok/s

Qwen3.5 122B

• Can RTX Spark N1X run Qwen3.5 122B? Yes: 35.2 tok/s

Qwen3.5 9B

• Can RTX 4090 run Qwen3.5 9B? Yes: 137 tok/s
• Can RTX 5090 run Qwen3.5 9B? Yes: 208 tok/s

Mistral Medium 3.5 128B

• Can 2x Apple M5 Ultra run Mistral Medium 3.5 128B? Yes: 13.69 tok/s
• Can 4x Apple M5 Ultra run Mistral Medium 3.5 128B? Yes: 23.37 tok/s
• Can Apple M3 Ultra run Mistral Medium 3.5 128B? Yes: 4.95 tok/s
• Can Apple M5 Ultra run Mistral Medium 3.5 128B? Yes: 7.7 tok/s

Muse Glimmer 30B

• Can RTX A4000 run Muse Glimmer 30B? Yes: 21.42 tok/s
• Can RTX Spark N1X run Muse Glimmer 30B? Yes: 15.9 tok/s

Nemotron 3 Nano Omni 30B

• Can DGX Spark (GB10) run Nemotron 3 Nano Omni 30B? Yes: 58.8 tok/s
• Can RTX Spark N1X run Nemotron 3 Nano Omni 30B? Yes: 63.9 tok/s

GLM 4.7 Flash

• Can RTX 3090 run GLM 4.7 Flash? Yes: 134.6 tok/s
• Can RTX 4090 run GLM 4.7 Flash? Yes: 164.9 tok/s
• Can RTX Spark N1X run GLM 4.7 Flash? Yes: 55.6 tok/s

LFM2.5 24B-A2B

• Can RTX 3060 12GB run LFM2.5 24B-A2B? Partly: part of the model runs from system RAM (CPU offload, per the source): 125 tok/s
• Can RTX 4060 Ti 16GB run LFM2.5 24B-A2B? Yes: 131-150 tok/s
• Can RTX 4090 run LFM2.5 24B-A2B? Yes: 261-339 tok/s

Nemotron 3.5 Bolt 74B-A9B

• Can DGX Spark (GB10) run Nemotron 3.5 Bolt 74B-A9B? Yes: 61.61 tok/s
• Can RTX 4090 run Nemotron 3.5 Bolt 74B-A9B? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 108 tok/s

Nemotron 3.5 Super VL 120B-A12B

• Can DGX Spark (GB10) run Nemotron 3.5 Super VL 120B-A12B? Yes: 38.07 tok/s

Qwen3.8 Flash-Next

• Can 2x Intel Arc Pro B70 run Qwen3.8 Flash-Next? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 58.56 tok/s
• Can 2x RTX 5070 Ti run Qwen3.8 Flash-Next? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 182 tok/s
• Can 2x RTX 5080 run Qwen3.8 Flash-Next? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 144.4 tok/s
• Can 2x RTX PRO 4000 Blackwell run Qwen3.8 Flash-Next? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 100 tok/s
• Can 2x Tesla V100 32GB run Qwen3.8 Flash-Next? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 121 tok/s
• Can DGX Spark (GB10) run Qwen3.8 Flash-Next? Yes: 26.8 tok/s
• Can RTX 5090 run Qwen3.8 Flash-Next? Partly: part of the model runs from system RAM (CPU offload, per the source): 159.2 tok/s

Flint 40B-A2.5B

• Can Apple M4 Pro run Flint 40B-A2.5B? Yes: 127.9 tok/s
• Can RTX 3090 run Flint 40B-A2.5B? Yes: 128.9 tok/s
• Can RTX 4090 run Flint 40B-A2.5B? Yes: 179.4 tok/s

Nemotron 3.5 Lightning 30B-A3B

• Can DGX Spark (GB10) run Nemotron 3.5 Lightning 30B-A3B? Yes: 94.4 tok/s

Nemotron 3 Super 120B-A12B

• Can DGX Spark (GB10) run Nemotron 3 Super 120B-A12B? Yes: 23.0 tok/s

DeepSeek V4 Flash

• Can 8x RTX 3090 run DeepSeek V4 Flash? Yes: 95.52 tok/s

DeepSeek V4.1 Flash

• Can 2x Tesla V100 32GB run DeepSeek V4.1 Flash? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 13 tok/s
• Can RTX 5090 run DeepSeek V4.1 Flash? Partly: part of the model runs from system RAM (CPU offload, per the source): 24.9 tok/s

GLM-5.3 Flash

• Can 2x DGX Spark (GB10) run GLM-5.3 Flash? Yes: 56.5 tok/s
• Can 4x DGX Spark (GB10) run GLM-5.3 Flash? Yes: 105.3 tok/s
• Can RTX 5090 run GLM-5.3 Flash? Partly: part of the model runs from system RAM (CPU offload, per the source): 30.1 tok/s

Ornith 1.5 122B

• Can RTX 5090 run Ornith 1.5 122B? Partly: part of the model runs from system RAM (CPU offload, per the source): 46.4 tok/s

GLM-5.2

• Can 4x DGX Spark (GB10) run GLM-5.2? Yes: 32.5 tok/s

Ornith 1.5 35B-A3B

• Can Intel Arc Pro B70 run Ornith 1.5 35B-A3B? Yes: 131.46 tok/s

Qwen3 235B-A22B

• Can 4x RTX 3090 run Qwen3 235B-A22B? Yes: 18 tok/s

DeepSeek R1 671B

• Can Apple M3 Ultra run DeepSeek R1 671B? Yes: 19.89 tok/s
• Can RTX 3090 run DeepSeek R1 671B? Probably only partly: the weights are larger than its memory, so part of the model likely ran from system RAM: 8.97 tok/s