RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

What AI models run on the RX 7900 GRE?

The RX 7900 GRE has 16 GB of VRAM (576 GB/s). About 15.2 GB is usable for a model. 194 of 285 open models we track fit at 8K context.

194models fit (8K ctx)
180still fit at 32K ctx
40Blargest that fits
576 GB/smemory bandwidth

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)
flint40B (1.3B act.)Q2_K14.8 GBno1229details
Maxwell V1 Flash37B (2.3B act.)Q2_K14.0 GBQ2_K677details
Qwen3.6 35B A3B36B (3.0B act.)Q2_K13.6 GBQ2_K512details
Ornith 1.5 35B A3B36B (3.0B act.)Q2_K13.2 GBQ2_K528details
Qwen3.5 35B A3B36B (3.0B act.)Q2_K13.6 GBQ2_K512details
Axion 1.0 35B35B (3.9B act.)Q2_K13.3 GBQ2_K394details
Qwen AgentWorld 35B A3B35B (3.0B act.)Q2_K13.1 GBQ2_K512details
Yi 1.5 34B Chat34BQ2_K14.3 GBno45details
Nous Hermes 2 Yi 34B34BQ2_K14.8 GBno43details
DeepSeek R1 Distill Qwen 32B33BQ2_K14.2 GBno47details
Qwen3 32B33BQ2_K14.1 GBno47details
GLM Z1 32B 041433BQ2_K12.6 GBQ2_K47details
GLM 4 32B 041433BQ2_K12.6 GBQ2_K47details
granite 4.0 h small32B (8.8B act.)Q2_K11.7 GBQ2_K179details
gemma 4 31B it31BQ2_K13.9 GBno49details
GLM 4.7 Flash31B (4.6B act.)Q2_K11.7 GBQ2_K344details
Qwen3 Coder 30B A3B Instruct31B (3.0B act.)Q2_K11.9 GBQ2_K521details
North Mini Code 1.030B (3.3B act.)Q2_K12.0 GBQ2_K465details
Hy Memory 30B A3B30B (3.0B act.)Q2_K11.9 GBQ2_K512details
Muse Glimmer 30B30BQ2_K11.4 GBQ2_K52details
granite 4.2 30b29BQ2_K12.7 GBno53details
Qwen3.8 27B28BQ2_K11.1 GBQ2_K55details
Qwen3.6 27B28BQ2_K11.1 GBQ2_K55details
Qwen3.5 27B28BQ2_K11.1 GBQ2_K55details
gemma 3 27b it27BQ2_K11.7 GBQ2_K55details
Arete 27B27BQ3_K_M13.6 GBQ2_K44details
MiniMax H3 Image PE27BQ2_K11.3 GBQ2_K54details
Trinity Mini26B (2.0B act.)Q3_K_M12.3 GBQ3_K_M633details
GLiDE 27B26BQ3_K_M13.2 GBQ2_K45details
diffusiongemma 26B A4B it26B (4.0B act.)Q3_K_M13.2 GBQ2_K295details
gemma 4 26B A4B it26B (4.0B act.)Q3_K_M13.3 GBQ2_K292details
LFM2.5 24B A2B24B (2.0B act.)Q3_K_M11.7 GBQ3_K_M589details
ERNIE 4.5 21B A3B PT22BQ4_K_M13.5 GBQ3_K_M43details
ERNIE 4.5 21B A3B Thinking22BQ4_K_M13.5 GBQ3_K_M43details
gpt oss safeguard 20b22B (4.8B act.)Q8_012.3 GBQ8_0214details
gpt oss 20b21B (4.2B act.)Q8_012.3 GBQ8_0237details
Groot 19B A3B 26102319B (3.0B act.)Q4_K_M12.7 GBQ4_K_M299details
Kimi VL A3B Thinking 250616B (3.0B act.)Q6_K13.5 GBQ5_K_M234details
Kimi VL A3B Instruct16B (3.0B act.)Q5_K_M12.2 GBQ5_K_M262details
Kimi VL A3B Thinking16B (3.0B act.)Q6_K13.5 GBQ5_K_M234details
deepseek moe 16b base16B (2.8B act.)Q4_K_M12.5 GBQ2_K307details
LLaDA2.1 mini16B (1.4B act.)Q6_K13.4 GBQ5_K_M490details
Moonlight 16B A3B Instruct16B (3.0B act.)Q5_K_M12.2 GBQ5_K_M255details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)Q5_K_M11.9 GBQ5_K_M287details
DeepSeek V2 Lite16B (2.7B act.)Q5_K_M11.3 GBQ5_K_M304details
DeepSeek V2 Lite Chat16B (2.7B act.)Q5_K_M11.9 GBQ5_K_M287details
Qwen2.5 Coder 14B Instruct15BQ6_K13.5 GBQ3_K_M48details
DeepSeek R1 Distill Qwen 14B15BQ6_K13.5 GBQ3_K_M48details
Qwen3 14B15BQ6_K13.2 GBQ3_K_M48details
phi 415BQ6_K13.4 GBQ3_K_M48details
gemma 3 12b it12BQ8_013.4 GBQ6_K46details
Llama Guard 4 12B12BQ6_K11.5 GBQ4_K_M59details
Llama 3.2 11B Vision Instruct11BQ8_011.6 GBQ5_K_M55details
Llama 3.2 11B Vision11BQ8_012.4 GBQ5_K_M51details
GLM 4.6V Flash10BQ8_010.3 GBQ8_058details
GLM 4.1V 9B Thinking10BQ8_010.3 GBQ8_058details
Step3 VL 10B10BQ8_09.9 GBQ8_066details
SenseNova Loop 9B Preview9.7BQ8_010.0 GBQ8_060details
Qwen3.5 9B9.7BQ8_010.0 GBQ8_060details
Ornith 1.5 9B9.7BQ8_010.0 GBQ8_060details
Qwen3.5 9B Base9.7BQ8_010.0 GBQ8_060details
MiMo V2.6 Distill Qwen 9B9.4BQ8_010.0 GBQ8_060details
glm 4 9b9.4BQ6_K13.4 GBno70details
internlm3 8b instruct8.8BQ8_09.7 GBQ8_062details
granite 4.2 8b8.8BQ8_010.5 GBQ6_K62details
granite 4.1 8b8.8BQ8_010.5 GBQ6_K62details
Qwen3 VL 8B Instruct8.8BQ8_09.9 GBQ8_066details
Molmo2 8B8.7BQ8_09.9 GBQ8_066details
LFM2.5 8B A1B8.5B (1.0B act.)Q8_09.1 GBQ8_0541details
Qwen2.5 VL 7B Instruct8.3BQ8_08.7 GBQ8_071details
UI TARS 1.5 7B8.3BQ8_08.7 GBQ8_071details
olmOCR 2 7B 10258.3BQ8_08.7 GBQ8_071details
olmOCR 7B 0225 preview8.3BQ8_08.7 GBQ8_071details
UI TARS 7B DPO8.3BQ8_08.7 GBQ8_071details
Qwen3 8B8.2BQ8_09.9 GBQ8_066details
DeepSeek R1 0528 Qwen3 8B8.2BQ8_09.9 GBQ8_066details
granite 3.3 8b instruct8.2BQ8_09.8 GBQ8_066details
granite 3.1 8b instruct8.2BQ8_09.8 GBQ8_066details
granite 3.2 8b instruct8.2BQ8_09.8 GBQ8_066details
Apertus 8B Instruct 25098.1BQ8_09.6 GBQ8_067details
Llama 3.1 8B Instruct8.0BQ8_09.6 GBQ8_068details
Llama 3.1 8B8.0BQ8_09.6 GBQ8_067details
Hermes 3 Llama 3.1 8B8.0BQ8_09.6 GBQ8_067details
DeepSeek R1 Distill Llama 8B8.0BQ8_09.6 GBQ8_067details
Meta Llama 3 8B Instruct8.0BQ8_09.6 GBQ8_067details
Meta Llama Guard 2 8B8.0BQ8_09.6 GBQ8_067details
Meta Llama 3.1 8B Instruct8.0BQ8_09.6 GBQ8_067details
Ling 3.0 tiny7.9B (1.4B act.)Q8_08.7 GBQ8_0391details
MiMo 7B RL7.8BQ8_09.4 GBQ8_071details
internlm2 5 7b chat7.7BQ8_09.2 GBQ8_070details
internlm2 chat 7b7.7BQ8_09.2 GBQ8_070details
Qwen2.5 7B Instruct7.6BQ8_08.7 GBQ8_071details
Qwen2.5 Coder 7B Instruct7.6BQ8_08.7 GBQ8_071details
DeepSeek R1 Distill Qwen 7B7.6BQ8_08.7 GBQ8_071details
Falcon H1R 7B7.6BQ8_08.5 GBQ8_071details
Olmo 3 7B Instruct7.3BQ8_010.3 GBQ8_074details
Olmo 3 1025 7B7.3BQ8_010.3 GBQ8_074details
Olmo 3 7B Think7.3BQ8_010.3 GBQ8_074details
Mistral 7B v0.17.2BQ8_08.6 GBQ8_075details
Mistral 7B Instruct v0.27.2BQ8_08.6 GBQ8_075details
Mistral 7B Instruct v0.17.2BQ8_08.6 GBQ8_075details
falcon 7b7.2BQ8_012.1 GBno75details
falcon 7b instruct7.2BQ8_011.8 GBno78details
granite 4.0 h tiny6.9B (1.5B act.)F1613.6 GBQ8_0196details
OLMoE 1B 7B 09246.9B (1.3B act.)Q8_08.4 GBQ8_0422details
deepseek coder 7b instruct v1.56.9BQ8_011.2 GBno78details
deepseek coder 6.7b instruct6.7BQ8_011.1 GBno80details
deepseek coder 6.7b base6.7BQ8_011.1 GBno80details
Llama 2 7b hf6.7BQ8_011.1 GBno80details
Yi 6B 200K6.1BF1612.3 GBQ8_048details
Yi 6B6.1BF1612.3 GBQ8_048details
Molmo2 4B4.9BF1610.0 GBF1665details
WeVisOne VL 4B4.8BF1610.0 GBF1665details
Mage VL Cowork4.7BF1610.0 GBF1665details
Qianfan OCR4.7BF1610.1 GBF1665details
Qwen3.5 4B4.7BF169.0 GBF1668details
S5 nano4.7BF169.2 GBF1666details
LightOnOCR 3 4B4.5BF169.0 GBF1668details
SingVL S1 4B4.5BF169.0 GBF1668details
Intern Decision 4B4.5BF169.0 GBF1668details
Qwen3 VL 4B Instruct4.4BF169.3 GBF1672details
gemma 3 4b it4.3BF168.4 GBF1674details
medgemma 1.5 4b it4.3BF168.4 GBF1674details
medgemma 4b it4.3BF168.4 GBF1674details
gemma 3 4b pt4.3BF169.2 GBF1667details
Phi 3.5 vision instruct4.1BF1611.2 GBQ4_K_M69details
Qwen3 4B4.0BF169.3 GBF1672details
Qwen3 4B Instruct 25074.0BF169.3 GBF1672details
granite vision 4.1 4b4.0BF167.6 GBF1685details
Phi 4 mini instruct3.8BF168.9 GBF1675details
Phi 3 mini 4k instruct3.8BF1610.6 GBQ5_K_M75details
Phi 3.5 mini instruct3.8BF1610.6 GBQ5_K_M75details
Phi 3 mini 128k instruct3.8BF1610.6 GBQ5_K_M75details
Qwen2.5 VL 3B Instruct3.8BF166.7 GBF1693details
granite 4.2 3b3.7BF168.0 GBF1679details
granite 4.1 3b llmon preview3.7BF168.0 GBF1679details
Raon 1 nano3.5BF167.0 GBF1692details
granite 4.1 3b3.4BF167.6 GBF1685details
granite 4.0 micro3.4BF167.6 GBF1685details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF16278details
deepseek vl2 tiny3.4B (1.2B act.)F167.4 GBF16245details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF16285details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF16287details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF16287details
granite 3.1 3b a800m instruct3.3B (883M act.)F167.1 GBF16326details
Llama 3.2 3B Instruct3.2BF1613.5 GBQ8_045details
Llama 3.2 3B3.2BF167.5 GBF1690details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF1690details
d1 3B3.1BF1610.8 GBF1653details
LFM2.5 VL 3B3.1BF1610.8 GBF1653details
Qwen2.5 3B Instruct3.1BF166.7 GBF1693details
SmolLM3 3B3.1BF166.9 GBF1694details
SmolLM3 3B Base3.1BF166.9 GBF1694details
phi 22.8BF168.2 GBQ8_0104details
LFM2.5 2.6B2.7BF1610.8 GBF1653details
granite 3.3 2b instruct2.5BF165.8 GBF16114details
MiniCPM5 2B2.5BF165.7 GBF16114details
WeVisOne VL 2B2.4BF165.4 GBF16141details
Qwen3.5 2B2.3BF164.5 GBF16153details
SmolVLM Instruct2.2BF165.4 GBF16159details
Qwen3 1.7B2.0BF164.8 GBF16167details
Qwen3.8 27B DSpine b162.0BF164.8 GBF16142details
d1 2B1.9BF167.2 GBF1683details
LFM3 VL 2B1.9BF167.2 GBF1683details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF16162details
LFM3 1.7B1.7BF167.2 GBF1683details
LFM3 1.7B Base1.7BF164.0 GBF16167details
SmolLM2 1.7B Instruct1.7BF165.2 GBF16168details
SmolLM2 1.7B1.7BF165.2 GBF16168details
SmolLM 1.7B1.7BF165.2 GBF16168details
granite 4.0 1b1.6BF164.3 GBF16176details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF16187details
OLMo 2 0425 1B1.5BF164.4 GBF16194details
deepseek coder 1.3b instruct1.3BF164.5 GBF16214details
Llama 3.2 1B1.2BF163.2 GBF16233details
Llama 3.2 1B Instruct1.2BF165.5 GBF16116details
LFM2.5 1.2B Instruct1.2BF165.0 GBF16123details
HunyuanOCR1.1BF163.5 GBF16257details
PaddleOCR VL 1.6959MF162.5 GBF16300details
Qwen3.5 0.8B873MF162.4 GBF16380details
LightOnOCR 3 0.8B853MF162.4 GBF16380details
FastVLM 0.5B759MF162.2 GBF16380details
Qwen3 0.6B752MF162.7 GBF16481details
GOT OCR2 0716MF162.8 GBF16402details
Nemotron 3.5 Parse631MF161.9 GBF16441details
GOT OCR 2.0 hf561MF162.5 GBF16514details
SmolVLM2 500M Video Instruct507MF161.6 GBF16702details
SmolVLM 500M Instruct507MF161.6 GBF16702details
Qwen2.5 0.5B Instruct494MF161.7 GBF16583details
LFM2.5 VL 450M449MF161.9 GBF16405details
SmolLM2 360M Instruct362MF161.5 GBF16794details
SmolLM2 360M362MF161.5 GBF16794details
LFM2.5 350M354MF161.9 GBF16405details
LFM2 350M354MF161.4 GBF16812details

Just too big at 4-bit

These need up to 35% more memory than the RX 7900 GRE has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
Trinity Mini15.8 GB0.6 GB
GLiDE 27B16.2 GB1.0 GB
MiniMax H3 Image PE16.8 GB1.6 GB
Qwen3.8 27B16.8 GB1.6 GB
Qwen3.5 27B17.0 GB1.8 GB
Qwen3.6 27B17.0 GB1.8 GB
diffusiongemma 26B A4B it17.1 GB1.9 GB
gemma 4 26B A4B it17.2 GB2.0 GB
gemma 3 27b it17.3 GB2.1 GB
Arete 27B17.4 GB2.2 GB
Muse Glimmer 30B17.9 GB2.7 GB
GLM 4.7 Flash18.2 GB3.0 GB
Hy Memory 30B A3B18.5 GB3.3 GB
Qwen3 Coder 30B A3B Instruct18.7 GB3.5 GB
granite 4.0 h small18.9 GB3.7 GB

Questions

What is the biggest model that runs on the RX 7900 GRE?

Of the models we track, the largest that fits at 8K context is flint (40B) at Q2_K, needing about 14.8 GB.

How much of the RX 7900 GRE's memory can a model use?

About 15.2 GB: we assume 95% of the 16 GB is usable after driver and display overhead.

Spec source: AMD specifications. Model data: Hugging Face Hub API. Speed is a bandwidth ceiling (bandwidth ÷ bytes read per token), not a benchmark. Methodology · Report an error.