AI Compatibility Tier List
AMD Instinct MI350X 288GB
374 AI models ranked by how well they run on your 288 GB VRAM hardware. Best match: DeepSeek V4 Flash (score: 98).
374
Models tested
342
Can run
288 GB
VRAM
126
Best tok/s
Tier Distribution
S 29
A 55
B 66
C 192
F 32
S: 29A: 55B: 66C: 192F: 32
S
S Tier -- Perfect Fit (29 models)
These models run flawlessly with headroom to spare.
DeepSeek V4 Flash284B189.0 GB126 tok/s1049K ctx
moe
Qwen 3.5 397B A17B397B274.7 GB79 tok/s90K ctx
moe
Qwen 3 235B A22B235B175.9 GB119 tok/s131K ctx
moe
Devstral 2 123B Instruct123B110.1 GB85 tok/s256K ctx
dense
DeepSeek Coder V2 236B236B232.3 GB109 tok/s31K ctx
moe
Qwen 3.5 122B A10B122B106.6 GB235 tok/s131K ctx
moe
MiniMax M2.7
S91230B173.8 GB135 tok/s205K ctx
moe
DeepSeek V2.5 236B236B232.3 GB109 tok/s31K ctx
moe
Mistral Small 4 119B119B107.7 GB255 tok/s256K ctx
moe
GPT-OSS 120B117B106.0 GB89 tok/s131K ctx
dense
Command A 111B111B101.3 GB94 tok/s262K ctx
dense
Pixtral Large 124B124B110.7 GB84 tok/s131K ctx
dense
Qwen3-Coder 30B A3B Instruct30.5B49.8 GB883 tok/s256K ctx
moe
Qwen 3.6 35B A3B35B55.2 GB742 tok/s262K ctx
moe
Qwen 2.5 VL 72B72B78.5 GB145 tok/s33K ctx
dense
Qwen 3.5 27B27B49.3 GB378 tok/s131K ctx
dense
Qwen 3.6 27B27B47.1 GB239 tok/s262K ctx
dense
Qwen3-VL 30B A3B Instruct30B49.5 GB913 tok/s256K ctx
moe
Llama 4 Maverick 17B 128E400B276.6 GB77 tok/s78K ctx
moe
Qwen3-Coder-Next80B80.0 GB394 tok/s256K ctx
moe
Qwen 3.5 35B A3B35B52.5 GB807 tok/s131K ctx
moe
Magistral Small 250724B46.8 GB336 tok/s131K ctx
dense
Qwen 3 32B32B53.1 GB325 tok/s131K ctx
dense
Devstral Small 2 24B Instruct24B46.8 GB336 tok/s256K ctx
dense
Qwen 3.5 9B9B37.4 GB126 tok/s131K ctx
dense
Qwen 3 30B A3B30.5B49.8 GB883 tok/s131K ctx
moe
Leanstral 119B A6B119B111.1 GB234 tok/s256K ctx
moe
Nemotron 3 Nano 30B30B50.4 GB343 tok/s131K ctx
dense
Qwen 3 14B14B40.7 GB196 tok/s131K ctx
dense
A
A Tier -- Great Fit (55 models)
Excellent performance with comfortable memory margins.
Devstral Small 1.124B46.8 GB336 tok/s131K ctx
dense
Phi-4-reasoning-plus 14B14.7B41.7 GB206 tok/s33K ctx
dense
Qwen 3 8B8B36.8 GB112 tok/s131K ctx
dense
GPT-OSS 20B21B45.0 GB1121 tok/s128K ctx
moe
Llama 3.1 405B405B284.4 GB26 tok/s23K ctx
dense
Gemma 4 31B30.7B63.1 GB204 tok/s256K ctx
dense
Nemotron Cascade 2 30B A3B30B50.9 GB903 tok/s262K ctx
moe
Qwen 3.5 4B4B34.3 GB56 tok/s131K ctx
dense
Llama 3.3 70B70B77.3 GB149 tok/s128K ctx
dense
EXAONE 4.0 32B32B53.1 GB323 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B48.7 GB948 tok/s256K ctx
moe
Ministral 3 14B14B40.7 GB196 tok/s262K ctx
multimodal
Phi-4 Mini Reasoning 4B3.8B33.5 GB53 tok/s131K ctx
dense
Nemotron Nano 8B8B36.5 GB112 tok/s131K ctx
dense
Codestral 2 25.0822B45.6 GB308 tok/s256K ctx
dense
Qwen 2.5 32B32B53.1 GB323 tok/s131K ctx
dense
Mistral Small 3.2 24B24B46.8 GB336 tok/s131K ctx
vision
Llama 3.1 70B70B77.3 GB149 tok/s128K ctx
dense
Qwen 2.5 72B72B78.5 GB145 tok/s131K ctx
dense
Kimi Linear 48B A3B48B59.9 GB200 tok/s1049K ctx
linear
Gemma 3 27B27B57.4 GB231 tok/s131K ctx
dense
LFM2 24B24B46.8 GB336 tok/s131K ctx
dense
Qwen3-Coder 480B A35B Instruct480B325.4 GB35 tok/s4K ctx
moe
CogVLM2 19B19B43.7 GB266 tok/s8K ctx
dense
OLMo 2 32B32B53.1 GB323 tok/s4K ctx
dense
Granite 4.1 30B30B51.9 GB343 tok/s131K ctx
dense
Mistral Small 24B24B46.8 GB336 tok/s33K ctx
dense
Llama 4 Scout 17B 16E109B99.1 GB223 tok/s1048K ctx
moe
InternVL2 8B8B36.5 GB112 tok/s8K ctx
dense
Mistral Small 3.1 24B24B46.8 GB336 tok/s131K ctx
dense
Phi-4 14B14B41.3 GB196 tok/s16K ctx
dense
Qwen 2.5 14B14B41.2 GB196 tok/s131K ctx
dense
Ministral 3 8B8B36.8 GB112 tok/s262K ctx
multimodal
MiniCPM-V 2.6 8B8B36.5 GB112 tok/s2K ctx
dense
Granite Code 20B20B45.1 GB280 tok/s8K ctx
dense
Gemma 3 12B12B41.9 GB168 tok/s131K ctx
dense
Qwen 3 4B4B34.3 GB56 tok/s33K ctx
dense
SQLCoder 7B7B35.9 GB98 tok/s8K ctx
dense
Nemotron Nano 9B v29B37.6 GB126 tok/s131K ctx
dense
Qwen 2.5 VL 7B7B34.8 GB98 tok/s33K ctx
dense
DeepSeek R1 Distill 70B70B77.3 GB149 tok/s131K ctx
dense
DeepSeek Coder V2 16B16B42.8 GB1425 tok/s131K ctx
moe
Qwen 2.5 Coder 32B32B53.1 GB323 tok/s131K ctx
dense
Jina Embeddings v30.57B32.8 GB8 tok/s8K ctx
dense
Granite Code 34B34B54.1 GB305 tok/s8K ctx
dense
BGE M30.57B32.0 GB8 tok/s8K ctx
dense
Magistral 7B7B35.9 GB98 tok/s8K ctx
dense
CodeGeeX 4 9B9B35.8 GB126 tok/s131K ctx
dense
Command R 35B35B53.5 GB298 tok/s131K ctx
dense
Gemma 4 E4B8B35.9 GB112 tok/s128K ctx
dense
OLMo 2 13B13B40.1 GB182 tok/s33K ctx
dense
DeepSeek R1 Distill 32B32B53.1 GB323 tok/s33K ctx
dense
StarCoder 15B15B55.1 GB210 tok/s8K ctx
dense
CodeLlama 13B Instruct13B49.8 GB182 tok/s16K ctx
dense
Qwen 2.5 7B7B34.8 GB98 tok/s131K ctx
dense
B
B Tier -- Good Fit (66 models)
Solid performance, may need quantization or reduced context.
Granite Code 8B8B36.5 GB112 tok/s8K ctx
dense
DevStral 7B7B35.9 GB98 tok/s8K ctx
dense
Codestral Mamba 7B7B34.5 GB98 tok/s262K ctx
state-space
DeepSeek R1 Distill 14B14B41.2 GB196 tok/s33K ctx
dense
Nemotron 70B70B77.3 GB149 tok/s131K ctx
dense
mxbai Embed Large0.34B31.9 GB5 tok/s1K ctx
dense
LLaVA 1.6 13B13B49.8 GB182 tok/s4K ctx
dense
Granite 4.1 8B8B37.0 GB112 tok/s131K ctx
dense
Pixtral 12B12B39.5 GB168 tok/s131K ctx
dense
CodeLlama 7B Instruct7B41.8 GB98 tok/s16K ctx
dense
StarCoder 7B7B41.3 GB98 tok/s8K ctx
dense
Snowflake Arctic Embed L0.34B31.9 GB5 tok/s1K ctx
dense
MPT-30B-Instruct30B74.7 GB276 tok/s8K ctx
dense
Qwen 2.5 Coder 3B3B33.7 GB42 tok/s131K ctx
dense
Falcon 40B Instruct40B60.3 GB225 tok/s8K ctx
dense
BGE Large EN v1.50.34B31.9 GB5 tok/s1K ctx
dense
Gemma 4 E2B5.1B33.3 GB71 tok/s128K ctx
dense
Nous Dolphin 13B13B51.3 GB182 tok/s16K ctx
dense
Command R+ 104B104B96.6 GB100 tok/s131K ctx
dense
WizardLM 13B13B49.8 GB182 tok/s8K ctx
dense
Vicuna 13B13B49.8 GB182 tok/s4K ctx
dense
Nous Hermes 1.09B47.4 GB126 tok/s16K ctx
dense
GLM-4 9B9B35.8 GB126 tok/s128K ctx
dense
Gemma 2 27B27B57.4 GB231 tok/s8K ctx
dense
Llama 3.1 8B8B36.5 GB112 tok/s128K ctx
dense
InternLM 7B7B41.8 GB98 tok/s8K ctx
dense
InternLM Chat 7B7B41.8 GB98 tok/s8K ctx
dense
Solar 7B7B36.9 GB98 tok/s8K ctx
dense
WizardMath 7B7B35.9 GB98 tok/s4K ctx
dense
OLMo 2 7B7B35.9 GB98 tok/s4K ctx
dense
Nomic Embed Text v1.50.14B30.5 GB2 tok/s8K ctx
dense
Qwen 2.5 Coder 7B7B34.8 GB98 tok/s131K ctx
dense
Mixtral 8x22B141B119.1 GB141 tok/s66K ctx
moe
Ministral 3 3B3B32.3 GB42 tok/s262K ctx
multimodal
Gemma 3 4B4B34.2 GB56 tok/s128K ctx
dense
Phi 4 Mini 4B4B33.6 GB56 tok/s128K ctx
dense
LLaVA 1.5 7B7B41.8 GB98 tok/s4K ctx
dense
Qwen 3.5 2B2B32.6 GB28 tok/s131K ctx
dense
Baichuan 13B13B51.3 GB182 tok/s8K ctx
dense
Mixtral 8x7B47B60.3 GB420 tok/s33K ctx
moe
Cerebras-GPT 13B13B48.8 GB182 tok/s131K ctx
dense
DeepSeek R1 Distill 7B7B34.8 GB98 tok/s33K ctx
dense
DeepSeek R1 Distill 8B8B36.5 GB112 tok/s33K ctx
dense
MPT-7B-Instruct7B41.8 GB98 tok/s8K ctx
dense
Baichuan 7B7B41.8 GB98 tok/s8K ctx
dense
Qwen 2.5 Math 72B72B78.5 GB145 tok/s4K ctx
dense
Qwen 2.5 3B3B33.7 GB42 tok/s131K ctx
dense
Samantha 7B7B35.9 GB98 tok/s4K ctx
dense
Falcon 7B Instruct7B34.1 GB98 tok/s8K ctx
dense
Phi 3 Mini 3.8B3.8B37.9 GB53 tok/s128K ctx
dense
Qwen 3 1.7B1.7B32.4 GB24 tok/s33K ctx
dense
All MiniLM L6 v20.02B30.0 GB2 tok/s0K ctx
dense
Gemma 2 9B9B40.3 GB126 tok/s8K ctx
dense
Qwen 2.5 Coder 14B14B41.2 GB196 tok/s131K ctx
dense
Llama 3.2 11B Vision11B38.4 GB154 tok/s16K ctx
vision
Granite 4.1 3B3B32.8 GB42 tok/s131K ctx
dense
Phi 3.5 Mini 4B4B38.0 GB56 tok/s128K ctx
dense
Yi 1.5 34B34B54.1 GB306 tok/s4K ctx
dense
Qwen 2.5 Coder 1.5B1.5B31.0 GB21 tok/s33K ctx
dense
Granite Code 3B3B34.0 GB42 tok/s8K ctx
dense
DeepSeek LLM 67B67B76.4 GB155 tok/s4K ctx
dense
Mistral Nemo 12B12B39.5 GB168 tok/s128K ctx
dense
Yi Coder 9B9B36.7 GB126 tok/s131K ctx
dense
Mistral 7B Instruct v0.37B35.9 GB98 tok/s8K ctx
dense
Phi 3 Medium 14B14B41.3 GB196 tok/s128K ctx
dense
Codestral 22B22B45.6 GB308 tok/s33K ctx
dense
C
C Tier -- Tight Fit (192 models)
Runs with offloading or significant compromises.
InternLM 20B20B62.4 GB280 tok/s8K ctx
dense
Llama 3.2 3B3B33.2 GB42 tok/s128K ctx
dense
Ministral 8B8B36.8 GB112 tok/s131K ctx
dense
K EXAONE 236B A23B
C54236B201.3 GB41 tok/s66K ctx
dense
Baichuan M3 235B
C54235B200.6 GB41 tok/s67K ctx
dense
Baichuan M3 235B i1
C54235B200.6 GB41 tok/s67K ctx
dense
Aya Expanse 32B32B51.7 GB325 tok/s8K ctx
dense
DeepSeek R1 1.5B1.5B31.0 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B30.7 GB15 tok/s4K ctx
dense
Qwen3.5 122B A10B122B103.8 GB91 tok/s222K ctx
dense
SmolLM3 3B3B33.5 GB42 tok/s128K ctx
dense
Solar Open 100B
C49100B102.4 GB96 tok/s269K ctx
dense
Solar Open 100B i1
C49100B102.4 GB96 tok/s269K ctx
dense
Granite 3.1 8B8B36.5 GB112 tok/s128K ctx
state-space
Yi 1.5 9B9B36.7 GB126 tok/s4K ctx
dense
Qwen 2.5 Math 7B7B34.8 GB98 tok/s4K ctx
dense
Llama 3.3 70B Instruct
C4870B80.6 GB137 tok/s421K ctx
dense
Qwen 2.5 1.5B1.5B31.0 GB21 tok/s131K ctx
dense
Solar Open 69B REAP i1
C4869B79.9 GB139 tok/s428K ctx
dense
stabilityai japanese stablelm instruct beta 70b
C4770B80.6 GB137 tok/s421K ctx
dense
BaichuanMed OCR 72B i1
C4772B82.1 GB133 tok/s407K ctx
dense
Yi 34B Chat34B54.1 GB306 tok/s200K ctx
dense
OpenChat 7B7B35.9 GB98 tok/s8K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C4748B64.6 GB200 tok/s651K ctx
dense
Qwen3.5 35B A3B35B55.2 GB274 tok/s924K ctx
dense
StarCoder2 15B15B41.7 GB210 tok/s16K ctx
dense
Qwen3.5 35B A3B
C4635B55.2 GB274 tok/s924K ctx
dense
Aya Expanse 8B8B36.5 GB112 tok/s8K ctx
dense
Gemma 3 1B1B30.7 GB14 tok/s33K ctx
dense
Qwen3.5 27B27B49.3 GB355 tok/s1223K ctx
dense
Hermes 4.3 36B
C4636B55.9 GB266 tok/s896K ctx
dense
gemma 3 27b it27B49.3 GB355 tok/s1223K ctx
dense
gemma 3 27b it
C4627B49.3 GB355 tok/s1223K ctx
dense
EXAONE 4.0 32B
C4632B53.0 GB299 tok/s1019K ctx
dense
aya expanse 32b heretic MPOA i1
C4632B53.0 GB299 tok/s1019K ctx
dense
Vicuna 7B7B41.8 GB98 tok/s4K ctx
dense
Mistral Small 3.2 24B Instruct 250624B47.2 GB336 tok/s1386K ctx
dense
Baichuan M2 32B Q4 K M
C4532B53.0 GB299 tok/s1019K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4524B47.2 GB336 tok/s1386K ctx
dense
Mistral Small 24B Instruct 2501
C4524B47.2 GB336 tok/s1386K ctx
dense
baichuan inc Baichuan M2 32B
C4532B53.0 GB299 tok/s1019K ctx
dense
StableLM 2 12B12B50.5 GB168 tok/s4K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4524B47.2 GB336 tok/s1386K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4524B47.2 GB336 tok/s1386K ctx
dense
OpenHermes 2.5 7B7B35.9 GB98 tok/s8K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4524B47.2 GB336 tok/s1386K ctx
dense
Dolphin 2.9 8B8B36.5 GB112 tok/s33K ctx
dense
Codestral 22B v0.1
C4522B45.7 GB308 tok/s1520K ctx
dense
Codestral 22B v0.1
C4522B45.7 GB308 tok/s1520K ctx
dense
Gemma 2 2B2B32.5 GB28 tok/s8K ctx
dense
Zephyr 7B Beta7B35.9 GB98 tok/s33K ctx
dense
Codestral 22B v0.1
C4522B45.7 GB308 tok/s1520K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4524B47.2 GB336 tok/s1386K ctx
dense
DeepSeek R1 Distill Qwen 14B14B39.9 GB196 tok/s2436K ctx
dense
Qwen 2.5 Coder 0.5B0.5B30.2 GB7 tok/s131K ctx
dense
Qwen3.5 9B9B36.2 GB126 tok/s3835K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C459B36.2 GB126 tok/s3835K ctx
dense
Codestral 22B v0.1 IMat
C4522B45.7 GB308 tok/s1520K ctx
dense
Meta Llama 3.1 8B Instruct
C458B35.5 GB112 tok/s4325K ctx
dense
gemma 3 12b it
C4512B38.4 GB168 tok/s2856K ctx
dense
Starling LM 7B7B35.9 GB98 tok/s8K ctx
dense
llava llama 3 8b v1 1
C458B35.5 GB112 tok/s4325K ctx
dense
Phi 4 reasoning vision 15B
C4515B40.6 GB210 tok/s2268K ctx
dense
Qwen3.5 9B
C459B36.2 GB126 tok/s3835K ctx
dense
StarCoder2 15B
C4515B40.6 GB210 tok/s2268K ctx
dense
Llama 2 7B Chat
C457B34.8 GB98 tok/s4955K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B35.5 GB112 tok/s4325K ctx
dense
internlm2 5 20b chat
C4520B44.2 GB280 tok/s1680K ctx
dense
Codestral 22B v0.1 i1
C4522B45.7 GB308 tok/s1520K ctx
dense
Codestral 21B Pruned i1
C4521B45.0 GB294 tok/s1596K ctx
dense
internlm2 limarp chat 20b
C4520B44.2 GB280 tok/s1680K ctx
dense
Mistral 7B Instruct v0.2
C457B34.8 GB98 tok/s4955K ctx
dense
NousResearch Hermes 4 14B
C4514B39.9 GB196 tok/s2436K ctx
dense
Meta Llama 3 8B Instruct
C458B35.5 GB112 tok/s4325K ctx
dense
DeepSeek R1 Distill Llama 8B8B35.5 GB112 tok/s4325K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B35.5 GB112 tok/s4325K ctx
dense
Llama 3 8B Instruct 32k v0.1
C458B35.5 GB112 tok/s4325K ctx
dense
Mistral 7B Instruct v0.3
C457B34.8 GB98 tok/s4955K ctx
dense
Meta Llama 3.1 8B Instruct
C458B35.5 GB112 tok/s4325K ctx
dense
vntl llama3 8b v2
C458B35.5 GB112 tok/s4325K ctx
dense
internlm2 math plus 20b i1
C4520B44.2 GB280 tok/s1680K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B35.5 GB112 tok/s4325K ctx
dense
Yi Coder 9B Chat
C459B36.2 GB126 tok/s3835K ctx
dense
glm 4 9b chat 1m
C459B36.2 GB126 tok/s3835K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4510.7B37.5 GB150 tok/s3213K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B35.5 GB112 tok/s4325K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4510.7B37.5 GB150 tok/s3213K ctx
dense
starcoder2 15b instruct v0.1
C4515B40.6 GB210 tok/s2268K ctx
dense
Codestral RAG 19B Pruned i1
C4519B43.5 GB266 tok/s1773K ctx
dense
starcoder2 15b i1
C4515B40.6 GB210 tok/s2268K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C458B35.5 GB112 tok/s4325K ctx
dense
Hermes 3 Llama 3.1 8B
C458B35.5 GB112 tok/s4325K ctx
dense
starcoder2 15b instruct v0.1
C4515B40.6 GB210 tok/s2268K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4414B39.9 GB196 tok/s2436K ctx
dense
Mistral 7B Instruct v0.3
C447B34.8 GB98 tok/s4955K ctx
dense
Yi 1.5 9B Chat
C449B36.2 GB126 tok/s3835K ctx
dense
internlm JanusCoder 14B
C4414B39.9 GB196 tok/s2436K ctx
dense
Hermes 2 Pro Llama 3 8B
C448B35.5 GB112 tok/s4325K ctx
dense
Hermes 2 Pro Mistral 7B
C447B34.8 GB98 tok/s4955K ctx
dense
Dolphin3.0 Llama3.1 8B
C448B35.5 GB112 tok/s4325K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4414B39.9 GB196 tok/s2436K ctx
dense
HelpingAI 15B i1
C4415B40.6 GB210 tok/s2268K ctx
dense
granite 8b code instruct 4k
C448B35.5 GB112 tok/s4325K ctx
dense
zephyr 7B beta
C447B34.8 GB98 tok/s4955K ctx
dense
dolphin 2.9.4 llama3.1 8b
C448B35.5 GB112 tok/s4325K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C447B34.8 GB98 tok/s4955K ctx
dense
Falcon H1R 7B
C447B34.8 GB98 tok/s4955K ctx
dense
openchat 3.6 8b 20240522 IMat
C448B35.5 GB112 tok/s4325K ctx
dense
falcon mamba 7b instruct Q4 K M
C447B34.8 GB98 tok/s4955K ctx
dense
solar finalised finetuned Model 10.7B i1
C4410.7B37.5 GB150 tok/s3213K ctx
dense
EXAONE 3.5 7.8B Instruct
C447.8B35.4 GB109 tok/s4438K ctx
dense
aya expanse 8b
C448B35.5 GB112 tok/s4325K ctx
dense
zephyr 7B alpha
C447B34.8 GB98 tok/s4955K ctx
dense
speechless zephyr code functionary 7b
C447B34.8 GB98 tok/s4955K ctx
dense
dolphin v2 8b abliterated i1
C448B35.5 GB112 tok/s4325K ctx
dense
EXAONE 3.5 7.8B Instruct
C447.8B35.4 GB109 tok/s4438K ctx
dense
StarCoder2 7B
C447B34.8 GB98 tok/s4955K ctx
dense
SOLAR 10.7B v1.0
C4410.7B37.5 GB150 tok/s3213K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C447.8B35.4 GB109 tok/s4438K ctx
dense
japanese stablelm instruct gamma 7B
C447B34.8 GB98 tok/s4955K ctx
dense
Yi 1.5 6B Chat
C446B34.1 GB84 tok/s5794K ctx
dense
Yi 9B Coder i1
C449B36.2 GB126 tok/s3835K ctx
dense
aya expanse 8b orthogonal heretic i1
C448B35.5 GB112 tok/s4325K ctx
dense
exaone 3.0 7.8b it
C447.8B35.4 GB109 tok/s4438K ctx
dense
Falcon H1 7B Instruct
C447B34.8 GB98 tok/s4955K ctx
dense
Falcon H1R 7B7B34.8 GB98 tok/s4955K ctx
dense
HelpingAI2 9B
C449B36.2 GB126 tok/s3835K ctx
dense
stablelm 2 zephyr 1 6b
C446B34.1 GB84 tok/s5794K ctx
dense
aya expanse 8b orthogonal heretic
C448B35.5 GB112 tok/s4325K ctx
dense
HelpingAI2.5 10B i1
C4410B37.0 GB140 tok/s3443K ctx
dense
Mamba Codestral 7B v0.1
C447B34.8 GB98 tok/s4955K ctx
dense
stabilityai japanese stablelm base gamma 7b
C447B34.8 GB98 tok/s4955K ctx
dense
Helply 10.2b chat i1
C4410.2B37.1 GB143 tok/s3374K ctx
dense
HelpingAI2 9B i1
C449B36.2 GB126 tok/s3835K ctx
dense
HelpingAI 9B 200k i1
C449B36.2 GB126 tok/s3835K ctx
dense
baichuan2 7b chat
C447B34.8 GB98 tok/s4955K ctx
dense
DiscoPOP zephyr 7b gemma
C447B34.8 GB98 tok/s4955K ctx
dense
HelpingAI 9B i1
C449B36.2 GB126 tok/s3835K ctx
dense
internlm2 5 1 8b chat i1
C448B35.5 GB112 tok/s4325K ctx
dense
starcoder2 7b
C447B34.8 GB98 tok/s4955K ctx
dense
internlm3 8b instruct abliterated i1
C448B35.5 GB112 tok/s4325K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C447B34.8 GB98 tok/s4955K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C447B34.8 GB98 tok/s4955K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C447B34.8 GB98 tok/s4955K ctx
dense
internlm2 math plus 7b IMat
C447B34.8 GB98 tok/s4955K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C447B34.8 GB98 tok/s4955K ctx
dense
Yi 1.5 6B Chat
C446B34.1 GB84 tok/s5794K ctx
dense
internlm2 5 7b chat i1
C447B34.8 GB98 tok/s4955K ctx
dense
Yi 1.5 6B6B34.3 GB84 tok/s4K ctx
dense
stablelm 2 1 6b chat imatrix
C446B34.1 GB84 tok/s5794K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C447B34.8 GB98 tok/s4955K ctx
dense
MD Judge v0 2 internlm2 7b i1
C447B34.8 GB98 tok/s4955K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C447B34.8 GB98 tok/s4955K ctx
dense
jointpreferences mistral 7b sft helpful
C447B34.8 GB98 tok/s4955K ctx
dense
zephyr 7b dpo full i1
C447B34.8 GB98 tok/s4955K ctx
dense
blossom v3 baichuan2 7b i1
C447B34.8 GB98 tok/s4955K ctx
dense
blossom v1 baichuan 7b i1
C447B34.8 GB98 tok/s4955K ctx
dense
Neural Chat 7B7B35.9 GB98 tok/s8K ctx
dense
HelpingAI2 6B
C446B34.1 GB84 tok/s5794K ctx
dense
StarCoder2 7B7B34.5 GB98 tok/s16K ctx
dense
HelpingAI2 6B i1
C446B34.1 GB84 tok/s5794K ctx
dense
Nemotron Mini 4B4B34.1 GB56 tok/s4K ctx
dense
DeepSeek LLM 7B7B41.3 GB98 tok/s4K ctx
dense
Qwen3.5 4B4B32.6 GB56 tok/s8733K ctx
dense
HelpingAI2.5 5B i1
C435B33.3 GB70 tok/s6970K ctx
dense
gemma 3 4b it
C434B32.6 GB56 tok/s8733K ctx
dense
Qwen 3 0.6B0.6B30.9 GB8 tok/s33K ctx
dense
gemma 3 4b it
C434B32.6 GB56 tok/s8733K ctx
dense
Llama 3.2 1B1B30.8 GB14 tok/s128K ctx
dense
Llama 3.2 3B Instruct
C423B32.2 GB42 tok/s11657K ctx
dense
Qwen2.5 3B Instruct3B31.9 GB42 tok/s11672K ctx
dense
Llama 3.2 3B Instruct
C423B31.9 GB42 tok/s11672K ctx
dense
Ministral 3 3B Instruct 2512
C423B31.9 GB42 tok/s11672K ctx
dense
HELVETE 3B
C423B31.9 GB42 tok/s11672K ctx
dense
Hermes 3 Llama 3.2 3B
C423B31.9 GB42 tok/s11672K ctx
dense
stablelm zephyr 3b
C423B31.9 GB42 tok/s11672K ctx
dense
StarCoder2 3B
C423B31.9 GB42 tok/s11672K ctx
dense
AI21 Jamba Reasoning 3B
C413B31.9 GB42 tok/s11672K ctx
dense
stablelm 3b 4e1t
C413B31.9 GB42 tok/s11672K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C413B31.9 GB42 tok/s11672K ctx
dense
ai21labs AI21 Jamba2 3B
C413B31.9 GB42 tok/s11672K ctx
dense
HelpingAI 3B hindi i1
C413B31.9 GB42 tok/s11672K ctx
dense
AI21 Jamba2 3B
C413B31.9 GB42 tok/s11672K ctx
dense
HelpingAI 3B hindi
C413B31.9 GB42 tok/s11672K ctx
dense
AI21 Jamba2 3B i1
C413B31.9 GB42 tok/s11672K ctx
dense
EXAONE 3.5 2.4B Instruct
C412.4B31.4 GB34 tok/s14611K ctx
dense
gemma 2 2b it
C412B31.6 GB28 tok/s17521K ctx
dense
gemma 2b2B31.2 GB28 tok/s17550K ctx
dense
Gemmasutra Mini 2B v1
C412B31.2 GB28 tok/s17550K ctx
dense
Qwen 3.5 0.6B0.6B30.9 GB8 tok/s131K ctx
dense
gemma 2 2b it
C402B31.2 GB28 tok/s17550K ctx
dense
StarCoder2 3B3B32.0 GB42 tok/s16K ctx
dense
F
F Tier -- Can't Run (32 models)
Too heavy for this hardware.
Qwen2.5 1.5B Instruct1.5B30.8 GB21 tok/s23428K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B30.8 GB21 tok/s23428K ctx
dense
Yi Coder 1.5B Chat
D401.5B30.8 GB21 tok/s23428K ctx
dense
stablelm 2 zephyr 1.6b
D391.6B30.9 GB22 tok/s21958K ctx
dense
logos16v2 stablelm2 1.6b i1
D391.6B30.9 GB22 tok/s21958K ctx
dense
Yi Coder 1.5B
D391.5B30.8 GB21 tok/s23428K ctx
dense
Falcon H1 1.5B Instruct1.5B30.8 GB21 tok/s23428K ctx
dense
TinyLlama 1.1B Chat v1.0
D391.1B30.5 GB15 tok/s31977K ctx
dense
Qwen 2.5 0.5B0.5B30.2 GB7 tok/s131K ctx
dense
EXAONE 4.0 1.2B
D391.2B30.6 GB17 tok/s29306K ctx
dense
Llama 3.2 1B Instruct Q8 0
D381B30.6 GB14 tok/s35155K ctx
dense
TinyLlama 1.1B Chat v0.3
D381.1B30.5 GB15 tok/s31977K ctx
dense
TinyLlama 1.1B Chat v0.61.1B30.5 GB15 tok/s31977K ctx
dense
Qwen3.5 397B A17B397B318.4 GB15 tok/s6K ctx
dense
Llama 3.2 1B Instruct
D381B30.4 GB14 tok/s35183K ctx
dense
gemma 3 1b it
D381B30.4 GB14 tok/s35183K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D381.1B30.5 GB15 tok/s31977K ctx
dense
Falcon3 1B Instruct abliterated
D381B30.4 GB14 tok/s35183K ctx
dense
SmolVLM 500M Instruct
D360.5B30.2 GB7 tok/s41262K ctx
dense
embeddinggemma 300M
D350.3B30.0 GB4 tok/s41289K ctx
dense
granite embedding 107m multilingual
D330.11B29.9 GB2 tok/s41318K ctx
dense
Falcon H1 Tiny 90M Instruct
D320.09B29.9 GB2 tok/s41319K ctx
dense
Mistral Large 3675B446.8 GB13 tok/s4K ctx
moe
GLM-5744B502.6 GB9 tok/s4K ctx
moe
GLM-5.1754B508.7 GB8 tok/s4K ctx
moe
Kimi K2.51000B647.1 GB5 tok/s4K ctx
moe
Kimi K2.61000B647.1 GB5 tok/s4K ctx
moe
DeepSeek V3 671B671B498.6 GB10 tok/s4K ctx
moe
DeepSeek V3.1 671B671B498.6 GB10 tok/s4K ctx
moe
DeepSeek V4 Pro1600B893.6 GB4 tok/s4K ctx
moe
DeepSeek R1 671B671B498.6 GB10 tok/s4K ctx
moe
DeepSeek V3.2671B439.5 GB14 tok/s4K ctx
moe
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs