AMD Radeon PRO W7900
48 GB · 864 GB/s · 295 W
In short
The AMD Radeon PRO W7900 has 48 GB of GDDR6 memory with 864 GB/s of bandwidth, which sets how fast it generates text. At 64k context it runs 77 of the 125 open models we track fully in memory; the most capable is Qwen 3.8 Flash Next at UD-Q3_K_XL, at about 9 tok/s (estimated). We have no current price for the AMD Radeon PRO W7900 in the US.
Updated . Speeds are estimates; how they are worked out.
Prices
Every retailer we track in the US, cheapest first. Buy links may earn us a commission.
No retailer in this market lists it right now.
What two of them run
Every model in the catalogue, with the best quant for this hardware. Estimates, not measurements.
| Model | Quant | Fits | Memory | Max context | Decode @64k | Prefill | Good for |
|---|---|---|---|---|---|---|---|
| Qwen 3.8 Flash Next Alibaba Qwen | UD-Q3_K_XL | fits | 87.1 GB | 256k | 9tok/sest. | 323 tok/s | – |
| Qwen 3.8 27B Alibaba Qwen | Q8_0 | fits | 32.9 GB | 256k | 14tok/sest. | 627 tok/s | chat |
| MiniMax M2.7 MiniMax | UD-Q2_K_XL | fits | 87.3 GB | 92k170k with q8 cache | 14tok/sest. | 596 tok/s | chat, coding |
| Qwen 3.5 122B-A10B Alibaba Qwen | Q5_K_M | fits | 88.5 GB | 256k | 31tok/sest. | 1305 tok/s | chat, coding, agents, long documents |
| Qwen 3.6 27B Alibaba Qwen | Q8_0 | fits | 32.4 GB | 256k | 14tok/sest. | 627 tok/s | chat |
| gpt-oss 120B OpenAI | MXFP4 | fits | 63.0 GB | 128k | 52tok/sest. | 2024 tok/s | chat, coding, agents, long documents |
| Qwen3 Coder Next Alibaba Qwen | Q8_0 | fits | 82.2 GB | 256k | 49tok/sest. | 3024 tok/s | chat, coding, agents, long documents |
| Nemotron 3 Super 120B-A12B NVIDIA | Q5_K_M | fits | 92.3 GB | 217k256k with q8 cache | 28tok/sest. | 1517 tok/s | chat, coding, long documents |
| Gemma 4 31B | Q8_0 | fits | 37.9 GB | 256k | 12tok/sest. | 572 tok/s | – |
| Mistral Small 4 119B Mistral AI | Q6_K | fits | 94.0 GB | 76k122k with q8 cache | 39tok/sest. | 1283 tok/s | chat, coding, agents, long documents |
| Qwen 3.6 35B-A3B Alibaba Qwen | Q8_0 | fits | 37.4 GB | 256k | 64tok/sest. | 3774 tok/s | chat, coding, agents, long documents |
| Qwen 3.5 27B Alibaba Qwen | Q8_0 | fits | 32.4 GB | 256k | 14tok/sest. | 627 tok/s | chat |
| Gemma 4 26B-A4B | Q8_0 | fits | 28.1 GB | 256k | 50tok/sest. | 3864 tok/s | chat, coding, agents, long documents |
| GLM 4.5 Air Z.ai | Q5_K_M | fits | 91.0 GB | 82k128k with q8 cache | 13tok/sest. | 425 tok/s | chat |
| Qwen 3.5 35B-A3B Alibaba Qwen | Q8_0 | fits | 37.4 GB | 256k | 64tok/sest. | 3774 tok/s | chat, coding, agents, long documents |
| Qwen3 Next 80B-A3B Alibaba Qwen | Q8_0 | fits | 82.2 GB | 256k | 59tok/sest. | 3449 tok/s | chat, coding, agents, long documents |
| GLM 4.7 Flash Z.ai | Q8_0 | fits | 34.6 GB | 198k | 32tok/sest. | 1872 tok/s | chat, coding, agents, long documents |
| Devstral Small 2 24B Mistral AI | Q8_0 | fits | 35.0 GB | 384k | 13tok/sest. | 617 tok/s | chat |
| Nemotron 3.5 Lightning 30B-A3B NVIDIA | UD-Q6_K_XL | fits | 34.7 GB | 256k | 76tok/sest. | 4653 tok/s | chat, coding, agents, long documents |
| Kimi Linear 48B-A3B Moonshot AI | Q8_0 | fits | 50.9 GB | 1024k | 76tok/sest. | 3032 tok/s | chat, coding, long documents |
| gpt-oss 20B OpenAI | MXFP4 | fits | 14.5 GB | 128k | 71tok/sest. | 2889 tok/s | chat, coding, agents, long documents |
| Qwen3 Coder 30B-A3B Alibaba Qwen | Q8_0 | fits | 37.9 GB | 256k | 30tok/sest. | 1351 tok/s | chat, coding, agents, long documents |
| Qwen3 32B Alibaba Qwen | Q8_0 | fits | 50.1 GB | 40k | 9tok/sest. | 320 tok/s | – |
| Gemma 4 12B | Q8_0 | fits | 14.8 GB | 256k | 34tok/sest. | 1486 tok/s | chat, coding, long documents |
| Qwen 3.5 9B Alibaba Qwen | Q8_0 | fits | 12.6 GB | 256k | 41tok/sest. | 1818 tok/s | chat, coding, long documents |
| Gemma 3 27B | Q8_0 | fits | 33.8 GB | 128k | 14tok/sest. | 702 tok/s | chat |
| Mistral Small 3.2 24B Mistral AI | Q8_0 | fits | 35.0 GB | 128k | 13tok/sest. | 617 tok/s | chat |
| DeepSeek R1 Distill Qwen 32B DeepSeek | Q8_0 | fits | 50.1 GB | 128k | 9tok/sest. | 396 tok/s | – |
| Llama 3.3 70B Meta | Q8_0 | fits | 91.5 GB | 73k128k with q8 cache | 5tok/sest. | 189 tok/s | – |
| Ministral 3 14B Mistral AI | Q8_0 | fits | 25.0 GB | 256k | 19tok/sest. | 869 tok/s | chat, coding |
| Qwen3 14B Alibaba Qwen | Q8_0 | fits | 26.3 GB | 40k | 18tok/sest. | 761 tok/s | chat, coding |
| DeepSeek R1 0528 Qwen3 8B DeepSeek | Q8_0 | fits | 18.8 GB | 128k | 26tok/sest. | 1202 tok/s | chat, coding, long documents |
| Qwen3 8B Alibaba Qwen | Q8_0 | fits | 18.8 GB | 40k | 26tok/sest. | 1202 tok/s | chat, coding, long documents |
| Phi-4 14B Microsoft | Q8_0 | fits | 28.7 GB | 16k | 16tok/sest. | 764 tok/s | chat, coding |
| Qwen 3.5 4B Alibaba Qwen | Q8_0 | fits | 7.9 GB | 256k | 71tok/sest. | 3152 tok/s | chat, coding, long documents |
| Ministral 3 8B Mistral AI | Q8_0 | fits | 18.6 GB | 256k | 26tok/sest. | 1189 tok/s | chat, coding, long documents |
| Granite 4.2 8B IBM | Q8_0 | fits | 20.4 GB | 128k | 24tok/sest. | 1099 tok/s | chat, coding, agents, long documents |
| Gemma 4 E4B | Q8_0 | fits | 11.1 GB | 128k | 47tok/sest. | 2382 tok/s | chat, coding, long documents |
| LFM2.5 8B-A1B Liquid AI | Q8_0 | fits | 13.1 GB | 125k | 67tok/sest. | 5083 tok/s | chat, coding, long documents |
| Llama 3.1 8B Meta | Q6_K | fits | 15.8 GB | 128k | 31tok/sest. | 1291 tok/s | chat, coding, long documents |
| Qwen2.5 0.5B Alibaba Qwen | Q8_0 | fits | 2.9 GB | 32k | 356tok/sest. | 11272 tok/s | chat, coding, long documents |
| Qwen3.5 0.8B Alibaba Qwen | Q8_0 | fits | 3.2 GB | 256k | 291tok/sest. | 12780 tok/s | chat, coding, agents, long documents |
| Qwen3.5 2B Alibaba Qwen | Q5_K_M | fits | 3.8 GB | 256k | 210tok/sest. | 6967 tok/s | chat, coding, agents, long documents |
| DeepSeek R1 Distill Qwen 1.5B DeepSeek | Q8_0 | fits | 5.2 GB | 128k | 126tok/sest. | 4669 tok/s | chat, coding, long documents |
| MiniCPM5 2B openbmb | Q8_0 | fits | 6.8 GB | 128k | 86tok/sest. | 2631 tok/s | chat, coding, agents, long documents |
| Spark X2.5 4B XHToken | mlx-4bit | fits | 6.8 GB | 1024k | 86tok/sest. | 3231 tok/s | chat, coding, agents, long documents |
| Mellum2.1 12B A2.5B Thinking JetBrains | Q8_0 | fits | 14.6 GB | 128k | 78tok/sest. | 4710 tok/s | chat, coding, agents, long documents |
| FrogNano 4B 2609 Microsoft | Q8_0 | fits | 8.0 GB | 256k | 70tok/sest. | 3152 tok/s | chat, coding, agents, long documents |
| LFM2.5 2.6B Liquid AI | Q8_0 | fits | 8.1 GB | 128k | 69tok/sest. | 3191 tok/s | chat, coding, agents, long documents |
| LFM2.5 VL 3B Liquid AI | Q8_0 | fits | 8.1 GB | 32k | 69tok/sest. | 3001 tok/s | chat, coding, agents, long documents |
| Kimi VL A3B Thinking 2506 Moonshot AI | Q8_0 | fits | 19.4 GB | 128k | 55tok/sest. | 3239 tok/s | chat, coding, long documents |
| Kimi VL A3B Moonshot AI | mlx-8bit | fits | 20.2 GB | 128k | 54tok/sest. | 3239 tok/s | chat, coding, agents, long documents |
| granite vision 4.1 4b IBM | Q8_0 | fits | 10.0 GB | 128k | 53tok/sest. | 2004 tok/s | chat, coding, long documents |
| granite 4.2 3b IBM | Q8_0 | fits | 10.3 GB | 128k | 51tok/sest. | 2069 tok/s | chat, coding, agents, long documents |
| Ministral 3 3B Instruct 2512 Mistral AI | Q8_0 | fits | 11.6 GB | 256k | 45tok/sest. | 1981 tok/s | chat, coding, agents, long documents |
| GLM 4 9B 0414 Z.ai | mlx-5bit | fits | 11.9 GB | 32k | 44tok/sest. | 1065 tok/s | chat, coding, agents, long documents |
| DeepSeek R1 Distill Qwen 7B DeepSeek | Q8_0 | fits | 12.7 GB | 128k | 40tok/sest. | 1512 tok/s | chat, coding, long documents |
| LFM2.5 24B A2B Liquid AI | Q8_0 | fits | 30.4 GB | 128k | 38tok/sest. | 2912 tok/s | chat, coding, agents, long documents |
| GLM 4.6V Flash Z.ai | Q8_0 | fits | 13.5 GB | 128k | 38tok/sest. | 1020 tok/s | chat, coding, agents, long documents |
| GLM 4.1V 9B Thinking Z.ai | mlx-8bit | fits | 15.1 GB | 64k | 33tok/sest. | 1020 tok/s | chat, coding, long documents |
| DeepSeek R1 Distill Llama 8B DeepSeek | Q8_0 | fits | 17.6 GB | 128k | 28tok/sest. | 1291 tok/s | chat, coding, long documents |
| Ministral 3 8B Reasoning 2512 Mistral AI | Q8_0 | fits | 18.6 GB | 256k | 26tok/sest. | 1189 tok/s | chat, coding, agents, long documents |
| Ministral 3 14B Reasoning 2512 Mistral AI | Q8_0 | fits | 25.0 GB | 256k | 19tok/sest. | 869 tok/s | chat, coding |
| Qwen3.8 27B OBLITERATED OBLITERATUS | mlx-4bit | fits | 25.1 GB | 256k | 19tok/sest. | 627 tok/s | chat, coding, long documents |
| DeepSeek R1 Distill Qwen 14B DeepSeek | Q8_0 | fits | 28.3 GB | 128k | 17tok/sest. | 695 tok/s | chat, coding |
| Muse Glimmer 30B meta-models | Q8_0 | fits | 32.7 GB | 128k | 14tok/sest. | 633 tok/s | – |
| GLM 4.5V Z.ai | Q4_K_M | fits | 72.4 GB | 64k | 14tok/sest. | 418 tok/s | chat |
| Devstral Small 2507 Mistral AI | Q8_0 | fits | 35.0 GB | 128k | 13tok/sest. | 625 tok/s | chat |
| Magistral Small 2506 Mistral AI | Q8_0 | fits | 35.0 GB | 40k | 13tok/sest. | 625 tok/s | chat |
| Magistral Small 2507 Mistral AI | Q8_0 | fits | 35.0 GB | 40k | 13tok/sest. | 625 tok/s | chat |
| Magistral Small 2509 Mistral AI | Q8_0 | fits | 35.0 GB | 128k | 13tok/sest. | 617 tok/s | chat |
| GLM 4.6V Z.ai | Q4_K_M | fits | 78.8 GB | 128k | 13tok/sest. | 418 tok/s | chat |
| granite 4.2 30b IBM | Q8_0 | fits | 46.7 GB | 128k | 10tok/sest. | 462 tok/s | – |
| DeepSeek R1 Distill Llama 70B DeepSeek | Q8_0 | fits | 91.5 GB | 73k128k with q8 cache | 5tok/sest. | 189 tok/s | – |
| Devstral 2 123B Instruct 2512 Mistral AI | Q4_K_M | fits | 93.4 GB | 66k123k with q8 cache | 5tok/sest. | 110 tok/s | – |
| Kimi Dev 72B Moonshot AI | Q8_0 | fits | 93.6 GB | 66k123k with q8 cache | 5tok/sest. | 186 tok/s | – |
| Mistral Medium 3.5 128B Mistral AI | Q4_K_M | fits | 93.4 GB | 66k123k with q8 cache | 5tok/sest. | 108 tok/s | – |
| DeepSeek V4 Flash DeepSeek | UD-Q2_K_XL | offload | 97.2 GB | 32k57k with q8 cache | 18tok/sest. | 187 tok/s | chat, coding |
| GLM 5.3 Flash Z.ai | UD-Q2_K_XL | offload | 103.7 GB | – | 18tok/sest. | 275 tok/s | chat |
| DeepSeek R1 DeepSeek | Q4_K_M | won't run | 382.6 GB | – | – | – | – |
| DeepSeek R1 0528 DeepSeek | Q4_K_M | won't run | 383.1 GB | – | – | – | – |
| DeepSeek R1 Zero DeepSeek | Q4_K_M | won't run | 382.6 GB | – | – | – | – |
| DeepSeek V3 DeepSeek | Q4_K_M | won't run | 382.6 GB | – | – | – | – |
| DeepSeek V3 0324 DeepSeek | Q4_K_M | won't run | 382.6 GB | – | – | – | – |
| DeepSeek V3.1 DeepSeek | Q4_K_M | won't run | 383.5 GB | – | – | – | – |
| DeepSeek V3.1 Terminus DeepSeek | Q4_K_M | won't run | 383.5 GB | – | – | – | – |
| DeepSeek V3.2 DeepSeek | Q4_K_M | won't run | 383.5 GB | – | – | – | – |
| DeepSeek V4 Flash 0731 DeepSeek | MXFP4 | won't run | 152.7 GB | – | – | – | – |
| DeepSeek V4 Flash Vision Exp DeepSeek | MXFP4 | won't run | 161.5 GB | – | – | – | – |
| GLM 4.5 Z.ai | Q4_K_M | won't run | 226.3 GB | – | – | – | – |
| GLM 4.6 Z.ai | Q4_K_M | won't run | 225.5 GB | – | – | – | – |
| GLM 4.7 Z.ai | Q4_K_M | won't run | 226.3 GB | – | – | – | – |
| GLM 5 Z.ai | MXFP4 | won't run | 389.5 GB | – | – | – | – |
| GLM 5.1 Z.ai | MXFP4 | won't run | 427.1 GB | – | – | – | – |
| GLM 5.2 Z.ai | UD-Q4_K_M | won't run | 441.0 GB | – | – | – | – |
| GLM 5.3 Z.ai | Q8_0 | won't run | 753.5 GB | – | – | – | – |
| Kimi K2 Moonshot AI | Q4_K_M | won't run | 584.1 GB | – | – | – | – |
| Kimi K2 Instruct 0905 Moonshot AI | Q4_K_M | won't run | 584.5 GB | – | – | – | – |
| Kimi K2 Thinking Moonshot AI | Q4_K_M | won't run | 584.5 GB | – | – | – | – |
| Kimi K2.5 Moonshot AI | Q4_K_M | won't run | 584.5 GB | – | – | – | – |
| MiniMax M2 MiniMax | Q4_K_M | won't run | 146.0 GB | – | – | – | – |
| MiniMax M2.1 MiniMax | Q4_K_M | won't run | 146.0 GB | – | – | – | – |
| MiniMax M2.5 MiniMax | MXFP4 | won't run | 132.4 GB | – | – | – | – |
| MiniMax M3 MiniMax | MXFP4 | won't run | 248.9 GB | – | – | – | – |
Specifications
From the manufacturer's published specifications; sources are linked where we have them.
- GPU
- Radeon PRO W7900
- Memory
- 48 GB GDDR6
- Memory bus
- 384-bit
- Memory bandwidth
- 864 GB/s
- FP16 compute
- 122.6 TFLOPS
- Software
- ROCm
- Board power
- 295 W
- PCIe generation
- Gen 4
- PCIe lanes
- x16
Compare with
Questions
- How much memory does the AMD Radeon PRO W7900 have for AI models?
- The AMD Radeon PRO W7900 has 48 GB of GDDR6 memory with 864 GB/s of bandwidth, which sets how fast it generates text.
- What is the best model the AMD Radeon PRO W7900 can run locally?
- Qwen 3.8 Flash Next (UD-Q3_K_XL) fits fully in its memory at 64k context and generates at about 9 tok/s, an estimate from its memory bandwidth. 77 of the 125 models we track fit at that context.
- Can the AMD Radeon PRO W7900 run DeepSeek V4 Flash?
- Not fully in memory: at UD-Q2_K_XL and 64k context DeepSeek V4 Flash needs about 97 GB, so part of it runs from system RAM and it slows to about 18 tok/s (estimated).
- How much does the AMD Radeon PRO W7900 cost in the US?
- We have no current price for the AMD Radeon PRO W7900 in the US.
Speeds on this page are for 64k context. See the best hardware for local AI at every budget.