Your rig

Radeon Pro W7800 · 32 GB

128 GB system RAM · RDNA 3 · 576 GB/s · 30.4 GB usable VRAM
Edit hardware Export as JSON

Context length

KV cache at f16, 8K tokens

Quality floor
Q8_0 — near-lossless Q6_K Q5_K_M Q4_K_M — recommended Q3_K_M Q2_K — damaged
KV cache type
System RAM (GB)
Runtime
Ollamallama.cppLM StudiovLLM

4 of 5 runtimes drive this device.

Allow CPU offload Include partial-offload fits

Offloaded layers run at 60 GB/s host bandwidth — expect single-digit tokens per second.

Which one should I actually run?

Best models for your Radeon Pro W7800

Updated 8 Oct 2026 1 new this month

What do you want to do? Each row is the best-ranked model on that use case's shortlist that fits at Q4_K_M. The ordering is editorial; the sizes and speeds are computed. How the shortlists work.

RecommendationModelWhyQuant · totalTok/s est.
Best reasoning
Your pick
Gemma 4 31B
Google · Apache 2.0
89% AIME — the strongest maths of any model under 32B. Q4_K_M · 20.2 GB 20 Runs great
Best overall Qwen3.8 27B
Alibaba · Apache 2.0
Best capability per gigabyte: a current-generation dense 27B with vision and 262K context. Q4_K_M · 16.7 GB 22 Runs great
Best for coding Qwen3.8 27B
Alibaba · Apache 2.0
Terminal-Bench 73, DeepSWE 42 — a generation ahead of anything else that fits 24 GB. Q4_K_M · 16.7 GB 22 Runs great
Best for writing Gemma 4 31B
Google · Apache 2.0
The cleanest, least verbose prose of the 2026 24–32 GB class. Q4_K_M · 20.2 GB 20 Runs great
Best vision Qwen3.8 27B
Alibaba · Apache 2.0
Image and video in, OSWorld 84 — the strongest local vision-language model at 24 GB. Q4_K_M · 16.7 GB 22 Runs great
Best for agents Qwen3.8 27B
Alibaba · Apache 2.0
Computer use and tool calling are what 3.8 was trained for (OSWorld 84). Q4_K_M · 16.7 GB 22 Runs great
Best translation Qwen3.8 27B
Alibaba · Apache 2.0
119 languages, and Qwen translations read as idiomatic rather than literal. Q4_K_M · 16.7 GB 22 Runs great
Fastest good model Qwen3.6 35B-A3B
Alibaba · Apache 2.0
3.3B active per token: the fastest model that still competes with dense 27Bs. Q4_K_M · 20.9 GB 72 Runs great
Best long context Qwen3.8 27B
Alibaba · Apache 2.0
262K native, and only 16 of 64 blocks keep a KV cache — long context is cheap here. Q4_K_M · 24.2 GB
at 128K context
22 Runs great

Everything that fits

Runs great

58 of 80 · fits with headroom for a longer prompt
ModelParamsQuantWeights +KV 8KTotal VRAM headroom Tok/sRuntime
Qwen3.6 35B-A3B 35.9B MoE / 3.3B act Q4_K_M 20.2 GB 0.16 GB 20.9 GB
9.5 GB free
72 Ollama Run
Ornith 1.5 35B-A3B 35.9B MoE / 3.3B act Q4_K_M 20.2 GB 0.16 GB 20.9 GB
9.5 GB free
72 Ollama Run
LLM-jp 4 33B Thinking 33.2B Q4_K_M 18.7 GB 2.00 GB 21.3 GB
9.1 GB free
19 llama.cpp Run
Qwen3 32B 32.8B Q4_K_M 18.4 GB 2.00 GB 21.0 GB
9.4 GB free
19 Ollama Run
DeepSeek-R1-Distill-Qwen 32B 32.8B Q4_K_M 18.4 GB 2.00 GB 21.0 GB
9.4 GB free
19 Ollama Run
Qwen2.5-Coder 32B 32.8B Q4_K_M 18.4 GB 2.00 GB 21.0 GB
9.4 GB free
19 Ollama Run
Olmo 3.1 32B Instruct 32.2B Q4_K_M 18.1 GB 1.25 GB 20.0 GB
10.4 GB free
19 Ollama Run
Nemotron 3.5 Lightning 30B-A3B 31.6B MoE / 3.2B act Q4_K_M 17.8 GB 0.05 GB 18.4 GB
12.0 GB free
75 Ollama Run
Gemma 4 31B 31.3B Q4_K_M 17.6 GB 2.03 GB 20.2 GB
10.2 GB free
20 Ollama Run
GLM-4.7-Flash 30B-A3B 31.2B MoE / 3B act Q4_K_M 17.5 GB 0.41 GB 18.6 GB
11.8 GB free
80 Ollama Run
Qwen3 30B-A3B 30.5B MoE / 3.3B act Q4_K_M 17.1 GB 0.75 GB 18.5 GB
11.9 GB free
72 Ollama Run
Qwen3 Coder 30B-A3B 30.5B MoE / 3.3B act Q4_K_M 17.1 GB 0.75 GB 18.5 GB
11.9 GB free
72 Ollama Run
Muse Glimmer 30B 29.8B Q4_K_M 16.8 GB 0.18 GB 17.5 GB
12.9 GB free
21 Ollama Run
Granite 4.1 30B 28.9B Q4_K_M 16.3 GB 2.00 GB 18.9 GB
11.5 GB free
21 Ollama Run
Qwen3.8 27B 27.8B Q4_K_M 15.6 GB 0.50 GB 16.7 GB
13.7 GB free
22 Ollama Run
Qwen3.6 27B 27.8B Q4_K_M 15.6 GB 0.50 GB 16.7 GB
13.7 GB free
22 Ollama Run
Gemma 3 27B 27.4B Q4_K_M 15.4 GB 1.03 GB 17.0 GB
13.4 GB free
23 Ollama Run
Gemma 4 26B-A4B 26.5B MoE / 3.8B act Q4_K_M 14.9 GB 0.51 GB 16.0 GB
14.4 GB free
63 Ollama Run
Devstral Small 2 24B 24B Q4_K_M 13.5 GB 1.25 GB 15.3 GB
15.1 GB free
26 llama.cpp Run
Mistral Small 3.2 24B 23.6B Q4_K_M 13.3 GB 1.25 GB 15.1 GB
15.3 GB free
26 Ollama Run
gpt-oss 20B 20.9B MoE / 3.6B act MXFP4 10.8 GB 0.19 GB 11.6 GB
18.8 GB free
72 Ollama Run
Qwen3 14B 14.8B Q4_K_M 8.3 GB 1.25 GB 10.2 GB
20.2 GB free
42 Ollama Run
DeepSeek-R1-Distill-Qwen 14B 14.8B Q4_K_M 8.3 GB 1.50 GB 10.4 GB
20.0 GB free
42 Ollama Run
Qwen2.5-Coder 14B 14.8B Q4_K_M 8.3 GB 1.50 GB 10.4 GB
20.0 GB free
42 Ollama Run
Phi-4 14B 14.7B Q4_K_M 8.3 GB 1.56 GB 10.4 GB
20.0 GB free
42 Ollama Run
Ministral 3 14B 13.9B Q4_K_M 7.8 GB 1.25 GB 9.7 GB
20.7 GB free
45 Ollama Run
Gemma 3 12B 12.2B Q4_K_M 6.9 GB 0.81 GB 8.3 GB
22.1 GB free
51 Ollama Run
Mistral NeMo 12B 12.2B Q4_K_M 6.9 GB 1.25 GB 8.7 GB
21.7 GB free
51 Ollama Run
Gemma 4 12B 12B Q4_K_M 6.7 GB 0.81 GB 8.2 GB
22.2 GB free
52 Ollama Run
Qwen3.5 9B 9.65B Q4_K_M 5.4 GB 0.25 GB 6.3 GB
24.1 GB free
64 Ollama Run
Ornith 1.5 9B 9.41B Q4_K_M 5.3 GB 0.25 GB 6.1 GB
24.3 GB free
66 Ollama Run
Ministral 3 8B 8.92B Q4_K_M 5.0 GB 1.06 GB 6.7 GB
23.7 GB free
70 Ollama Run
Granite 4.1 8B 8.79B Q4_K_M 4.9 GB 1.25 GB 6.8 GB
23.6 GB free
71 Ollama Run
LFM2.5 8B-A1B 8.47B MoE / 1.5B act Q4_K_M 4.8 GB 0.09 GB 5.5 GB
24.9 GB free
159 Ollama Run
Fara 7B 8.29B Q4_K_M 4.7 GB 0.44 GB 5.7 GB
24.7 GB free
75 llama.cpp Run
Qwen3 8B 8.19B Q4_K_M 4.6 GB 1.13 GB 6.3 GB
24.1 GB free
76 Ollama Run
Llama 3.1 8B Instruct 8.03B Q4_K_M 4.5 GB 1.00 GB 6.1 GB
24.3 GB free
77 Ollama Run
Gemma 4 E4B 8.0B Q4_K_M 4.5 GB 0.14 GB 5.2 GB
25.2 GB free
78 Ollama Run
Ling 3.0 Tiny 7.9B-A1.3B 7.9B MoE / 1.3B act Q4_K_M 4.4 GB 0.05 GB 5.1 GB
25.3 GB free
183 llama.cpp Run
DeepSeek-R1-Distill-Qwen 7B 7.62B Q4_K_M 4.3 GB 0.44 GB 5.3 GB
25.1 GB free
81 Ollama Run
Qwen2.5-Coder 7B 7.62B Q4_K_M 4.3 GB 0.44 GB 5.3 GB
25.1 GB free
81 Ollama Run
Olmo 3 7B Instruct 7.3B Q4_K_M 4.1 GB 2.50 GB 7.2 GB
23.2 GB free
85 Ollama Run
Mistral 7B Instruct v0.3 7.25B Q4_K_M 4.1 GB 1.00 GB 5.7 GB
24.7 GB free
86 Ollama Run
Gemma 4 E2B 5.1B Q4_K_M 2.9 GB 0.07 GB 3.5 GB
26.9 GB free
122 Ollama Run
Qwen3.5 4B 4.66B Q4_K_M 2.6 GB 0.25 GB 3.5 GB
26.9 GB free
133 Ollama Run
Gemma 3 4B 4.3B Q4_K_M 2.4 GB 0.30 GB 3.3 GB
27.1 GB free
144 Ollama Run
Qwen3 4B 4.02B Q4_K_M 2.3 GB 1.13 GB 4.0 GB
26.4 GB free
154 Ollama Run
Ministral 3 3B 3.85B Q4_K_M 2.2 GB 0.81 GB 3.6 GB
26.8 GB free
161 Ollama Run
Phi-4-mini 3.8B 3.84B Q4_K_M 2.2 GB 1.00 GB 3.8 GB
26.6 GB free
161 Ollama Run
Granite 4.1 3B 3.4B Q4_K_M 1.9 GB 0.63 GB 3.1 GB
27.3 GB free
182 Ollama Run
Llama 3.2 3B Instruct 3.21B Q4_K_M 1.8 GB 0.88 GB 3.3 GB
27.1 GB free
193 Ollama Run
LFM2.5 2.6B 2.7B Q4_K_M 1.5 GB 0.13 GB 2.2 GB
28.2 GB free
230 llama.cpp Run
Qwen3.5 2B 2.27B Q4_K_M 1.3 GB 0.09 GB 2.0 GB
28.4 GB free
273 Ollama Run
Qwen3 1.7B 1.72B Q4_K_M 1.0 GB 0.88 GB 2.4 GB
28.0 GB free
361 Ollama Run
Llama 3.2 1B Instruct 1.24B Q4_K_M 0.7 GB 0.25 GB 1.5 GB
28.9 GB free
500 Ollama Run
Gemma 3 1B 1.0B Q4_K_M 0.6 GB 0.04 GB 1.2 GB
29.2 GB free
620 Ollama Run
Qwen3.5 0.8B 0.87B Q4_K_M 0.5 GB 0.09 GB 1.2 GB
29.2 GB free
713 Ollama Run
Qwen3 0.6B 0.6B Q4_K_M 0.3 GB 0.88 GB 1.8 GB
28.6 GB free
1034 Ollama Run

Won't fit in VRAM

22 models · some are possible with CPU offload
ModelParamsQuantWeights +KV 8KTotal Over budget Tok/sRuntime
Qwen3 235B-A22B 235B MoE / 22B act Q4_K_M 132.1 GB 1.47 GB 134.2 GB +103.8 GB ~1.4 llama.cpp Why
Qwen3.8-Flash-Next 180B-A6B 180B MoE / 6B act Q4_K_M 101.2 GB 0.19 GB 102.0 GB +71.6 GB ~5.6 llama.cpp Why
Qwen3.5 122B-A10B 125B MoE / 10B act Q4_K_M 70.3 GB 0.19 GB 71.1 GB +40.7 GB ~4.0 llama.cpp Why
Nemotron 3 Super 120B-A12B 124B MoE / 12B act Q4_K_M 69.7 GB 0.06 GB 70.4 GB +40.0 GB ~3.4 llama.cpp Why
Ling 3.0 Flash 124B-A5B 124B MoE / 5.1B act Q4_K_M 69.7 GB 0.06 GB 70.4 GB +40.0 GB ~7.9 llama.cpp Why
Mistral Small 4 119B-A6B 119B MoE / 6.5B act Q4_K_M 66.9 GB 0.18 GB 67.7 GB +37.3 GB ~6.3 llama.cpp Why
gpt-oss 120B 117B MoE / 5.1B act MXFP4 60.5 GB 0.29 GB 61.4 GB +31.0 GB ~9.4 llama.cpp Why
Llama 4 Scout 109B-A17B 109B MoE / 17B act Q4_K_M 61.3 GB 1.50 GB 63.4 GB +33.0 GB ~2.5 llama.cpp Why
Llama 3.3 70B Instruct 70.6B Q4_K_M 39.7 GB 2.50 GB 42.8 GB +12.4 GB ~2.4 llama.cpp Why
DeepSeek-R1-Distill-Llama 70B 70.6B Q4_K_M 39.7 GB 2.50 GB 42.8 GB +12.4 GB ~2.4 llama.cpp Why
Kimi K3 2.8T-A104B 2780B MoE / 104B act Q4_K_M 1563.2 GB 0.21 GB 1564.0 GB +1533.6 GB ~0.2 llama.cpp Why
Qwen3.8 2.4T-A95B 2446B MoE / 95B act Q4_K_M 1375.4 GB 0.72 GB 1376.7 GB +1346.3 GB ~0.3 llama.cpp Why
DeepSeek V4 Pro 1.6T-A49B 1650B MoE / 49B act Q4_K_M 927.8 GB 0.54 GB 928.9 GB +898.5 GB ~0.5 llama.cpp Why
Kimi K2.6 1T-A32B 1027B MoE / 32B act Q4_K_M 577.5 GB 0.54 GB 578.6 GB +548.2 GB ~0.8 llama.cpp Why
GLM-5.2 744B-A40B 753B MoE / 40B act Q4_K_M 423.4 GB 0.69 GB 424.7 GB +394.3 GB ~0.7 llama.cpp Why
GLM-5.3 744B-A40B 753B MoE / 40B act Q4_K_M 423.4 GB 0.69 GB 424.7 GB +394.3 GB ~0.7 llama.cpp Why
DeepSeek-R1 671B 671B MoE / 37B act Q4_K_M 377.3 GB 0.54 GB 378.4 GB +348.0 GB ~0.7 llama.cpp Why
DeepSeek V4.1 Flash 552B-A16B New 552B MoE / 16B act Q4_K_M 310.4 GB 0.04 GB 311.0 GB +280.6 GB ~1.7 — Why
Qwen3.5 397B-A17B 403B MoE / 17B act Q4_K_M 226.6 GB 0.23 GB 227.4 GB +197.0 GB ~1.7 llama.cpp Why
Ornith 1.5 397B-A17B 397B MoE / 17B act Q4_K_M 223.2 GB 0.23 GB 224.1 GB +193.7 GB ~1.7 llama.cpp Why
GLM-5.3-Flash 320B-A18B 321B MoE / 18B act Q4_K_M 180.5 GB 0.09 GB 181.2 GB +150.8 GB ~1.6 llama.cpp Why
DeepSeek V4 Flash 284B-A13B 284B MoE / 13B act Q4_K_M 159.7 GB 0.38 GB 160.7 GB +130.3 GB ~2.3 llama.cpp Why

Totals = quantised weights + f16 KV cache at 8K tokens + 0.6 GB runtime overhead, against 30.4 GB usable VRAM (32 GB less display and driver reserve). Tokens per second are estimated from 576 GB/s peak memory bandwidth at batch 1 — not measured. Read the method.


Same memory, different speed

These devices have the same 30.4 GB usable, so they run exactly the models above. Generation speed scales with memory bandwidth.

DeviceBandwidthSpeed vs yours
GeForce RTX 5090
Blackwell · 2025
1792 GB/s 3.11×
RTX 5000 Ada
Ada Lovelace · 2023
576 GB/s same

Best LLM for 32 GB VRAM, with every card in the class →

What more memory unlocks

At Q4_K_M and 8K context this device fits 58 of 80 models. Each step up, and the largest models it adds:

Step upModels fitAdds, largest first
48 GB card 60 (+2) Llama 3.3 70B Instruct, DeepSeek-R1-Distill-Llama 70B

Every device ranked by what it can run →