AI Compatibility Tier List
NVIDIA A100 40GB
374 AI models ranked by how well they run on your 40 GB VRAM hardware. Best match: Qwen 3.6 35B A3B (score: 99).
374
Models tested
309
Can run
40 GB
VRAM
166
Best tok/s
Tier Distribution
S 32
A 49
B 53
C 175
F 65
S: 32A: 49B: 53C: 175F: 65
S
S Tier -- Perfect Fit (32 models)
These models run flawlessly with headroom to spare.
Qwen 3.6 35B A3B35B30.4 GB166 tok/s54K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B25.0 GB198 tok/s180K ctx
moe
Qwen 3.5 35B A3B35B27.7 GB181 tok/s131K ctx
moe
Qwen 3.5 27B27B24.5 GB86 tok/s94K ctx
dense
Qwen3-VL 30B A3B Instruct30B24.7 GB204 tok/s184K ctx
moe
Qwen 3 32B32B28.3 GB73 tok/s64K ctx
dense
Qwen 3 30B A3B30.5B25.0 GB198 tok/s131K ctx
moe
Nemotron 3 Nano 30B30B25.6 GB77 tok/s110K ctx
dense
Magistral Small 250724B22.0 GB96 tok/s131K ctx
dense
Devstral Small 2 24B Instruct24B22.0 GB96 tok/s134K ctx
dense
Qwen 3.6 27B27B22.3 GB53 tok/s262K ctx
dense
Nemotron Cascade 2 30B A3B30B26.1 GB202 tok/s92K ctx
moe
Devstral Small 1.124B22.0 GB96 tok/s131K ctx
dense
GPT-OSS 20B21B20.2 GB251 tok/s128K ctx
moe
EXAONE 4.0 32B32B28.3 GB72 tok/s64K ctx
dense
Qwen 3 14B14B15.9 GB165 tok/s131K ctx
dense
Phi-4-reasoning-plus 14B14.7B16.9 GB157 tok/s33K ctx
dense
Qwen 3.5 9B9B12.6 GB126 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B23.9 GB212 tok/s86K ctx
moe
Qwen 2.5 32B32B28.3 GB72 tok/s64K ctx
dense
OLMo 2 32B32B28.3 GB72 tok/s4K ctx
dense
Gemma 4 31B30.7B38.3 GB46 tok/s18K ctx
dense
Qwen 3 8B8B12.0 GB112 tok/s131K ctx
dense
Granite 4.1 30B30B27.1 GB77 tok/s69K ctx
dense
Mistral Small 3.2 24B24B22.0 GB96 tok/s131K ctx
vision
Gemma 3 27B27B32.6 GB52 tok/s27K ctx
dense
Codestral 2 25.0822B20.8 GB93 tok/s142K ctx
dense
LFM2 24B24B22.0 GB96 tok/s131K ctx
dense
Mistral Small 24B24B22.0 GB96 tok/s33K ctx
dense
Qwen 3.5 4B4B9.5 GB56 tok/s131K ctx
dense
Ministral 3 14B14B15.9 GB164 tok/s174K ctx
multimodal
CogVLM2 19B19B18.9 GB121 tok/s8K ctx
dense
A
A Tier -- Great Fit (49 models)
Excellent performance with comfortable memory margins.
Mistral Small 3.1 24B24B22.0 GB96 tok/s131K ctx
dense
Qwen 2.5 Coder 32B32B28.3 GB72 tok/s64K ctx
dense
Kimi Linear 48B A3B48B35.1 GB45 tok/s100K ctx
linear
Nemotron Nano 8B8B11.7 GB112 tok/s131K ctx
dense
Granite Code 34B34B29.3 GB68 tok/s8K ctx
dense
Granite Code 20B20B20.3 GB116 tok/s8K ctx
dense
Phi-4 14B14B16.5 GB164 tok/s16K ctx
dense
Command R 35B35B28.7 GB67 tok/s90K ctx
dense
StarCoder 15B15B30.3 GB123 tok/s8K ctx
dense
Phi-4 Mini Reasoning 4B3.8B8.7 GB53 tok/s131K ctx
dense
DeepSeek R1 Distill 32B32B28.3 GB72 tok/s33K ctx
dense
Qwen 2.5 14B14B16.4 GB165 tok/s131K ctx
dense
Gemma 3 12B12B17.1 GB116 tok/s91K ctx
dense
InternVL2 8B8B11.7 GB112 tok/s8K ctx
dense
CodeLlama 13B Instruct13B25.0 GB165 tok/s16K ctx
dense
DeepSeek Coder V2 16B16B18.0 GB319 tok/s123K ctx
moe
Ministral 3 8B8B12.0 GB112 tok/s220K ctx
multimodal
MiniCPM-V 2.6 8B8B11.7 GB112 tok/s2K ctx
dense
LLaVA 1.6 13B13B25.0 GB165 tok/s4K ctx
dense
Nemotron Nano 9B v29B12.8 GB126 tok/s131K ctx
dense
SQLCoder 7B7B11.1 GB98 tok/s8K ctx
dense
OLMo 2 13B13B15.3 GB178 tok/s33K ctx
dense
Qwen 2.5 VL 7B7B10.0 GB98 tok/s33K ctx
dense
Qwen 3 4B4B9.5 GB56 tok/s33K ctx
dense
Nous Dolphin 13B13B26.5 GB142 tok/s16K ctx
dense
WizardLM 13B13B25.0 GB165 tok/s8K ctx
dense
Vicuna 13B13B25.0 GB165 tok/s4K ctx
dense
Magistral 7B7B11.1 GB98 tok/s8K ctx
dense
CodeGeeX 4 9B9B11.0 GB126 tok/s131K ctx
dense
Jina Embeddings v30.57B8.0 GB8 tok/s8K ctx
dense
Gemma 4 E4B8B11.1 GB112 tok/s128K ctx
dense
DeepSeek R1 Distill 14B14B16.4 GB165 tok/s33K ctx
dense
Gemma 2 27B27B32.6 GB52 tok/s8K ctx
dense
Nous Hermes 1.09B22.6 GB126 tok/s16K ctx
dense
CodeLlama 7B Instruct7B17.0 GB98 tok/s16K ctx
dense
StarCoder 7B7B16.5 GB98 tok/s8K ctx
dense
BGE M30.57B7.2 GB8 tok/s8K ctx
dense
Pixtral 12B12B14.7 GB168 tok/s131K ctx
dense
Granite Code 8B8B11.7 GB112 tok/s8K ctx
dense
Qwen 2.5 7B7B10.0 GB98 tok/s131K ctx
dense
DevStral 7B7B11.1 GB98 tok/s8K ctx
dense
Granite 4.1 8B8B12.2 GB112 tok/s131K ctx
dense
Falcon 40B Instruct40B35.5 GB50 tok/s8K ctx
dense
Baichuan 13B13B26.5 GB142 tok/s8K ctx
dense
Codestral Mamba 7B7B9.7 GB98 tok/s262K ctx
state-space
InternLM 7B7B17.0 GB98 tok/s8K ctx
dense
InternLM Chat 7B7B17.0 GB98 tok/s8K ctx
dense
Cerebras-GPT 13B13B24.0 GB142 tok/s42K ctx
dense
mxbai Embed Large0.34B7.1 GB5 tok/s1K ctx
dense
B
B Tier -- Good Fit (53 models)
Solid performance, may need quantization or reduced context.
Qwen 2.5 Coder 3B3B8.9 GB42 tok/s131K ctx
dense
Snowflake Arctic Embed L0.34B7.1 GB5 tok/s1K ctx
dense
Mixtral 8x7B47B35.5 GB94 tok/s33K ctx
moe
Llama 3.1 8B8B11.7 GB112 tok/s128K ctx
dense
GLM-4 9B9B11.0 GB126 tok/s128K ctx
dense
Solar 7B7B12.1 GB98 tok/s8K ctx
dense
LLaVA 1.5 7B7B17.0 GB98 tok/s4K ctx
dense
Yi 1.5 34B34B29.3 GB68 tok/s4K ctx
dense
Gemma 4 E2B5.1B8.5 GB71 tok/s128K ctx
dense
WizardMath 7B7B11.1 GB98 tok/s4K ctx
dense
OLMo 2 7B7B11.1 GB98 tok/s4K ctx
dense
BGE Large EN v1.50.34B7.1 GB5 tok/s1K ctx
dense
Qwen 2.5 Coder 7B7B10.0 GB98 tok/s131K ctx
dense
MPT-7B-Instruct7B17.0 GB98 tok/s8K ctx
dense
Baichuan 7B7B17.0 GB98 tok/s8K ctx
dense
Gemma 3 4B4B9.4 GB56 tok/s128K ctx
dense
Nomic Embed Text v1.50.14B5.7 GB2 tok/s8K ctx
dense
Ministral 3 3B3B7.5 GB42 tok/s262K ctx
multimodal
Phi 4 Mini 4B4B8.8 GB56 tok/s128K ctx
dense
Qwen 2.5 Coder 14B14B16.4 GB165 tok/s131K ctx
dense
DeepSeek R1 Distill 8B8B11.7 GB112 tok/s33K ctx
dense
Gemma 2 9B9B15.5 GB126 tok/s8K ctx
dense
Qwen 3.5 2B2B7.8 GB28 tok/s131K ctx
dense
DeepSeek R1 Distill 7B7B10.0 GB98 tok/s33K ctx
dense
Phi 3 Mini 3.8B3.8B13.1 GB53 tok/s90K ctx
dense
Samantha 7B7B11.1 GB98 tok/s4K ctx
dense
Llama 3.2 11B Vision11B13.6 GB154 tok/s16K ctx
vision
Codestral 22B22B20.8 GB99 tok/s33K ctx
dense
Phi 3.5 Mini 4B4B13.2 GB56 tok/s89K ctx
dense
Falcon 7B Instruct7B9.3 GB98 tok/s8K ctx
dense
Qwen 2.5 3B3B8.9 GB42 tok/s131K ctx
dense
Mistral Nemo 12B12B14.7 GB168 tok/s128K ctx
dense
InternLM 20B20B37.6 GB93 tok/s8K ctx
dense
Phi 3 Medium 14B14B16.5 GB164 tok/s128K ctx
dense
Qwen 3 1.7B1.7B7.6 GB24 tok/s33K ctx
dense
Aya Expanse 32B32B26.9 GB73 tok/s8K ctx
dense
Granite 4.1 3B3B8.0 GB42 tok/s131K ctx
dense
Yi Coder 9B9B11.9 GB126 tok/s131K ctx
dense
Granite Code 3B3B9.2 GB42 tok/s8K ctx
dense
All MiniLM L6 v20.02B5.2 GB2 tok/s0K ctx
dense
Mistral 7B Instruct v0.37B11.1 GB98 tok/s8K ctx
dense
Qwen 2.5 Coder 1.5B1.5B6.2 GB21 tok/s33K ctx
dense
Ministral 8B8B12.0 GB112 tok/s131K ctx
dense
Yi 34B Chat34B29.3 GB68 tok/s63K ctx
dense
Llama 3.2 3B3B8.4 GB42 tok/s128K ctx
dense
Qwen3.5 35B A3B35B30.4 GB61 tok/s54K ctx
dense
Qwen3.5 35B A3B
B5635B30.4 GB61 tok/s54K ctx
dense
EXAONE 4.0 32B
B5632B28.2 GB67 tok/s66K ctx
dense
aya expanse 32b heretic MPOA i1
B5632B28.2 GB67 tok/s66K ctx
dense
Baichuan M2 32B Q4 K M
B5632B28.2 GB67 tok/s66K ctx
dense
baichuan inc Baichuan M2 32B
B5532B28.2 GB67 tok/s66K ctx
dense
StableLM 2 12B12B25.7 GB116 tok/s4K ctx
dense
Hermes 4.3 36B
B5536B31.1 GB60 tok/s50K ctx
dense
C
C Tier -- Tight Fit (175 models)
Runs with offloading or significant compromises.
Qwen3.5 27B27B24.5 GB79 tok/s94K ctx
dense
gemma 3 27b it27B24.5 GB79 tok/s94K ctx
dense
gemma 3 27b it
C5527B24.5 GB79 tok/s94K ctx
dense
Mistral Small 3.2 24B Instruct 250624B22.4 GB89 tok/s116K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C5424B22.4 GB89 tok/s116K ctx
dense
Mistral Small 24B Instruct 2501
C5424B22.4 GB89 tok/s116K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C5424B22.4 GB89 tok/s116K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C5424B22.4 GB89 tok/s116K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C5424B22.4 GB89 tok/s116K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C5324B22.4 GB89 tok/s116K ctx
dense
Codestral 22B v0.1
C5322B20.9 GB97 tok/s135K ctx
dense
Codestral 22B v0.1
C5322B20.9 GB97 tok/s135K ctx
dense
Codestral 22B v0.1
C5322B20.9 GB97 tok/s135K ctx
dense
Granite 3.1 8B8B11.7 GB112 tok/s128K ctx
state-space
Yi 1.5 9B9B11.9 GB126 tok/s4K ctx
dense
Codestral 22B v0.1 IMat
C5322B20.9 GB97 tok/s135K ctx
dense
Codestral 22B v0.1 i1
C5322B20.9 GB97 tok/s135K ctx
dense
StarCoder2 15B15B16.9 GB135 tok/s16K ctx
dense
Codestral 21B Pruned i1
C5221B20.2 GB102 tok/s145K ctx
dense
internlm2 5 20b chat
C5220B19.4 GB107 tok/s156K ctx
dense
internlm2 limarp chat 20b
C5220B19.4 GB107 tok/s156K ctx
dense
SmolLM3 3B3B8.7 GB42 tok/s128K ctx
dense
internlm2 math plus 20b i1
C5220B19.4 GB107 tok/s156K ctx
dense
Vicuna 7B7B17.0 GB98 tok/s4K ctx
dense
Qwen 2.5 Math 7B7B10.0 GB98 tok/s4K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C5148B39.8 GB45 tok/s17K ctx
dense
Codestral RAG 19B Pruned i1
C5119B18.7 GB113 tok/s169K ctx
dense
DeepSeek R1 1.5B1.5B6.2 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B5.9 GB15 tok/s4K ctx
dense
Phi 4 reasoning vision 15B
C5015B15.8 GB143 tok/s236K ctx
dense
StarCoder2 15B
C5015B15.8 GB143 tok/s236K ctx
dense
DeepSeek R1 Distill Qwen 14B14B15.1 GB153 tok/s259K ctx
dense
starcoder2 15b instruct v0.1
C5015B15.8 GB143 tok/s236K ctx
dense
starcoder2 15b i1
C5015B15.8 GB143 tok/s236K ctx
dense
starcoder2 15b instruct v0.1
C5015B15.8 GB143 tok/s236K ctx
dense
HelpingAI 15B i1
C5015B15.8 GB143 tok/s236K ctx
dense
NousResearch Hermes 4 14B
C5014B15.1 GB153 tok/s259K ctx
dense
OpenChat 7B7B11.1 GB98 tok/s8K ctx
dense
Aya Expanse 8B8B11.7 GB112 tok/s8K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C5014B15.1 GB153 tok/s259K ctx
dense
internlm JanusCoder 14B
C5014B15.1 GB153 tok/s259K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C5014B15.1 GB153 tok/s259K ctx
dense
gemma 3 12b it
C4912B13.6 GB168 tok/s316K ctx
dense
DeepSeek LLM 7B7B16.5 GB98 tok/s4K ctx
dense
Dolphin 2.9 8B8B11.7 GB112 tok/s33K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4910.7B12.7 GB150 tok/s365K ctx
dense
Qwen 2.5 1.5B1.5B6.2 GB21 tok/s131K ctx
dense
OpenHermes 2.5 7B7B11.1 GB98 tok/s8K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4910.7B12.7 GB150 tok/s365K ctx
dense
Zephyr 7B Beta7B11.1 GB98 tok/s33K ctx
dense
Qwen3.5 9B9B11.4 GB126 tok/s449K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C489B11.4 GB126 tok/s449K ctx
dense
solar finalised finetuned Model 10.7B i1
C4810.7B12.7 GB150 tok/s365K ctx
dense
SOLAR 10.7B v1.0
C4810.7B12.7 GB150 tok/s365K ctx
dense
Qwen3.5 9B
C489B11.4 GB126 tok/s449K ctx
dense
Starling LM 7B7B11.1 GB98 tok/s8K ctx
dense
Yi Coder 9B Chat
C489B11.4 GB126 tok/s449K ctx
dense
glm 4 9b chat 1m
C489B11.4 GB126 tok/s449K ctx
dense
Meta Llama 3.1 8B Instruct
C488B10.7 GB112 tok/s516K ctx
dense
Helply 10.2b chat i1
C4810.2B12.3 GB143 tok/s387K ctx
dense
llava llama 3 8b v1 1
C488B10.7 GB112 tok/s516K ctx
dense
HelpingAI2.5 10B i1
C4810B12.2 GB140 tok/s396K ctx
dense
Yi 1.5 9B Chat
C489B11.4 GB126 tok/s449K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B10.7 GB112 tok/s516K ctx
dense
Meta Llama 3 8B Instruct
C488B10.7 GB112 tok/s516K ctx
dense
DeepSeek R1 Distill Llama 8B8B10.7 GB112 tok/s516K ctx
dense
Dolphin3.0 Llama3.1 8B
C488B10.7 GB112 tok/s516K ctx
dense
Llama 3 8B Instruct 32k v0.1
C488B10.7 GB112 tok/s516K ctx
dense
Meta Llama 3.1 8B Instruct
C488B10.7 GB112 tok/s516K ctx
dense
vntl llama3 8b v2
C488B10.7 GB112 tok/s516K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C488B10.7 GB112 tok/s516K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C488B10.7 GB112 tok/s516K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C488B10.7 GB112 tok/s516K ctx
dense
Yi 9B Coder i1
C489B11.4 GB126 tok/s449K ctx
dense
Hermes 3 Llama 3.1 8B
C488B10.7 GB112 tok/s516K ctx
dense
HelpingAI2 9B
C489B11.4 GB126 tok/s449K ctx
dense
Llama 2 7B Chat
C487B10.0 GB98 tok/s601K ctx
dense
Hermes 2 Pro Llama 3 8B
C488B10.7 GB112 tok/s516K ctx
dense
Dolphin3.0 Llama3.1 8B
C488B10.7 GB112 tok/s516K ctx
dense
granite 8b code instruct 4k
C488B10.7 GB112 tok/s516K ctx
dense
dolphin 2.9.4 llama3.1 8b
C478B10.7 GB112 tok/s516K ctx
dense
HelpingAI2 9B i1
C479B11.4 GB126 tok/s449K ctx
dense
Mistral 7B Instruct v0.2
C477B10.0 GB98 tok/s601K ctx
dense
HelpingAI 9B 200k i1
C479B11.4 GB126 tok/s449K ctx
dense
Mistral 7B Instruct v0.3
C477B10.0 GB98 tok/s601K ctx
dense
HelpingAI 9B i1
C479B11.4 GB126 tok/s449K ctx
dense
openchat 3.6 8b 20240522 IMat
C478B10.7 GB112 tok/s516K ctx
dense
aya expanse 8b
C478B10.7 GB112 tok/s516K ctx
dense
dolphin v2 8b abliterated i1
C478B10.7 GB112 tok/s516K ctx
dense
EXAONE 3.5 7.8B Instruct
C477.8B10.6 GB109 tok/s531K ctx
dense
EXAONE 3.5 7.8B Instruct
C477.8B10.6 GB109 tok/s531K ctx
dense
aya expanse 8b orthogonal heretic i1
C478B10.7 GB112 tok/s516K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C477.8B10.6 GB109 tok/s531K ctx
dense
exaone 3.0 7.8b it
C477.8B10.6 GB109 tok/s531K ctx
dense
Mistral 7B Instruct v0.3
C477B10.0 GB98 tok/s601K ctx
dense
Hermes 2 Pro Mistral 7B
C477B10.0 GB98 tok/s601K ctx
dense
aya expanse 8b orthogonal heretic
C478B10.7 GB112 tok/s516K ctx
dense
zephyr 7B beta
C477B10.0 GB98 tok/s601K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C477B10.0 GB98 tok/s601K ctx
dense
Falcon H1R 7B
C477B10.0 GB98 tok/s601K ctx
dense
internlm2 5 1 8b chat i1
C478B10.7 GB112 tok/s516K ctx
dense
internlm3 8b instruct abliterated i1
C478B10.7 GB112 tok/s516K ctx
dense
falcon mamba 7b instruct Q4 K M
C477B10.0 GB98 tok/s601K ctx
dense
Neural Chat 7B7B11.1 GB98 tok/s8K ctx
dense
zephyr 7B alpha
C477B10.0 GB98 tok/s601K ctx
dense
speechless zephyr code functionary 7b
C477B10.0 GB98 tok/s601K ctx
dense
StarCoder2 7B
C477B10.0 GB98 tok/s601K ctx
dense
Gemma 3 1B1B5.9 GB14 tok/s33K ctx
dense
japanese stablelm instruct gamma 7B
C477B10.0 GB98 tok/s601K ctx
dense
Falcon H1 7B Instruct
C477B10.0 GB98 tok/s601K ctx
dense
Falcon H1R 7B7B10.0 GB98 tok/s601K ctx
dense
Gemma 2 2B2B7.7 GB28 tok/s8K ctx
dense
Mamba Codestral 7B v0.1
C477B10.0 GB98 tok/s601K ctx
dense
stabilityai japanese stablelm base gamma 7b
C477B10.0 GB98 tok/s601K ctx
dense
baichuan2 7b chat
C477B10.0 GB98 tok/s601K ctx
dense
DiscoPOP zephyr 7b gemma
C477B10.0 GB98 tok/s601K ctx
dense
starcoder2 7b
C477B10.0 GB98 tok/s601K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C477B10.0 GB98 tok/s601K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C477B10.0 GB98 tok/s601K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C477B10.0 GB98 tok/s601K ctx
dense
internlm2 math plus 7b IMat
C477B10.0 GB98 tok/s601K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C477B10.0 GB98 tok/s601K ctx
dense
internlm2 5 7b chat i1
C477B10.0 GB98 tok/s601K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C477B10.0 GB98 tok/s601K ctx
dense
MD Judge v0 2 internlm2 7b i1
C477B10.0 GB98 tok/s601K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C477B10.0 GB98 tok/s601K ctx
dense
jointpreferences mistral 7b sft helpful
C477B10.0 GB98 tok/s601K ctx
dense
zephyr 7b dpo full i1
C477B10.0 GB98 tok/s601K ctx
dense
blossom v3 baichuan2 7b i1
C477B10.0 GB98 tok/s601K ctx
dense
blossom v1 baichuan 7b i1
C477B10.0 GB98 tok/s601K ctx
dense
Yi 1.5 6B Chat
C476B9.3 GB84 tok/s715K ctx
dense
stablelm 2 zephyr 1 6b
C476B9.3 GB84 tok/s715K ctx
dense
Yi 1.5 6B6B9.5 GB84 tok/s4K ctx
dense
Yi 1.5 6B Chat
C466B9.3 GB84 tok/s715K ctx
dense
stablelm 2 1 6b chat imatrix
C466B9.3 GB84 tok/s715K ctx
dense
StarCoder2 7B7B9.7 GB98 tok/s16K ctx
dense
HelpingAI2 6B
C466B9.3 GB84 tok/s715K ctx
dense
HelpingAI2 6B i1
C466B9.3 GB84 tok/s715K ctx
dense
Nemotron Mini 4B4B9.3 GB56 tok/s4K ctx
dense
Qwen 2.5 Coder 0.5B0.5B5.4 GB7 tok/s131K ctx
dense
HelpingAI2.5 5B i1
C455B8.5 GB70 tok/s875K ctx
dense
Qwen3.5 4B4B7.8 GB56 tok/s1115K ctx
dense
gemma 3 4b it
C454B7.8 GB56 tok/s1115K ctx
dense
gemma 3 4b it
C454B7.8 GB56 tok/s1115K ctx
dense
Qwen 3 0.6B0.6B6.1 GB8 tok/s33K ctx
dense
Llama 3.2 1B1B6.0 GB14 tok/s128K ctx
dense
Llama 3.2 3B Instruct
C443B7.4 GB42 tok/s1499K ctx
dense
Qwen2.5 3B Instruct3B7.1 GB42 tok/s1514K ctx
dense
Llama 3.2 3B Instruct
C433B7.1 GB42 tok/s1514K ctx
dense
Ministral 3 3B Instruct 2512
C433B7.1 GB42 tok/s1514K ctx
dense
HELVETE 3B
C433B7.1 GB42 tok/s1514K ctx
dense
Hermes 3 Llama 3.2 3B
C433B7.1 GB42 tok/s1514K ctx
dense
stablelm zephyr 3b
C433B7.1 GB42 tok/s1514K ctx
dense
StarCoder2 3B
C433B7.1 GB42 tok/s1514K ctx
dense
AI21 Jamba Reasoning 3B
C433B7.1 GB42 tok/s1514K ctx
dense
stablelm 3b 4e1t
C433B7.1 GB42 tok/s1514K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C433B7.1 GB42 tok/s1514K ctx
dense
ai21labs AI21 Jamba2 3B
C433B7.1 GB42 tok/s1514K ctx
dense
HelpingAI 3B hindi i1
C433B7.1 GB42 tok/s1514K ctx
dense
AI21 Jamba2 3B
C433B7.1 GB42 tok/s1514K ctx
dense
HelpingAI 3B hindi
C433B7.1 GB42 tok/s1514K ctx
dense
AI21 Jamba2 3B i1
C433B7.1 GB42 tok/s1514K ctx
dense
gemma 2 2b it
C422B6.8 GB28 tok/s2284K ctx
dense
EXAONE 3.5 2.4B Instruct
C422.4B6.6 GB34 tok/s1914K ctx
dense
gemma 2b2B6.4 GB28 tok/s2313K ctx
dense
StarCoder2 3B3B7.2 GB42 tok/s16K ctx
dense
Gemmasutra Mini 2B v1
C422B6.4 GB28 tok/s2313K ctx
dense
gemma 2 2b it
C422B6.4 GB28 tok/s2313K ctx
dense
Qwen 3.5 0.6B0.6B6.1 GB8 tok/s131K ctx
dense
Qwen2.5 1.5B Instruct1.5B6.0 GB21 tok/s3112K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B6.0 GB21 tok/s3112K ctx
dense
Yi Coder 1.5B Chat
C401.5B6.0 GB21 tok/s3112K ctx
dense
stablelm 2 zephyr 1.6b
C401.6B6.1 GB22 tok/s2912K ctx
dense
logos16v2 stablelm2 1.6b i1
C401.6B6.1 GB22 tok/s2912K ctx
dense
Yi Coder 1.5B
C401.5B6.0 GB21 tok/s3112K ctx
dense
F
F Tier -- Can't Run (65 models)
Too heavy for this hardware.
Falcon H1 1.5B Instruct1.5B6.0 GB21 tok/s3112K ctx
dense
TinyLlama 1.1B Chat v1.0
D391.1B5.7 GB15 tok/s4273K ctx
dense
EXAONE 4.0 1.2B
D391.2B5.8 GB17 tok/s3910K ctx
dense
Qwen 2.5 0.5B0.5B5.4 GB7 tok/s131K ctx
dense
Llama 3.2 1B Instruct Q8 0
D391B5.8 GB14 tok/s4680K ctx
dense
TinyLlama 1.1B Chat v0.3
D391.1B5.7 GB15 tok/s4273K ctx
dense
TinyLlama 1.1B Chat v0.61.1B5.7 GB15 tok/s4273K ctx
dense
Llama 3.2 1B Instruct
D391B5.6 GB14 tok/s4709K ctx
dense
gemma 3 1b it
D391B5.6 GB14 tok/s4709K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D391.1B5.7 GB15 tok/s4273K ctx
dense
Falcon3 1B Instruct abliterated
D391B5.6 GB14 tok/s4709K ctx
dense
SmolVLM 500M Instruct
D370.5B5.4 GB7 tok/s5550K ctx
dense
embeddinggemma 300M
D350.3B5.2 GB4 tok/s5577K ctx
dense
granite embedding 107m multilingual
D330.11B5.1 GB2 tok/s5606K ctx
dense
Falcon H1 Tiny 90M Instruct
D330.09B5.1 GB2 tok/s5607K ctx
dense
Llama 3.1 70B70B52.5 GB14 tok/s4K ctx
dense
Llama 3.3 70B70B52.5 GB14 tok/s4K ctx
dense
Qwen 2.5 72B72B53.7 GB13 tok/s4K ctx
dense
Qwen 3 235B A22B235B151.1 GB4 tok/s4K ctx
moe
Qwen3-Coder-Next80B55.2 GB34 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B300.6 GB2 tok/s4K ctx
moe
Qwen 3.5 122B A10B122B81.8 GB9 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B249.9 GB3 tok/s4K ctx
moe
Mixtral 8x22B141B94.3 GB5 tok/s4K ctx
moe
MPT-30B-Instruct30B49.9 GB29 tok/s8K ctx
dense
Devstral 2 123B Instruct123B85.3 GB3 tok/s4K ctx
dense
Mistral Small 4 119B119B82.9 GB9 tok/s4K ctx
moe
Leanstral 119B A6B119B86.3 GB8 tok/s4K ctx
moe
Mistral Large 3675B422.0 GB2 tok/s4K ctx
moe
GLM-5744B477.8 GB2 tok/s4K ctx
moe
GLM-5.1754B483.9 GB2 tok/s4K ctx
moe
Kimi K2.51000B622.3 GB2 tok/s4K ctx
moe
Kimi K2.61000B622.3 GB2 tok/s4K ctx
moe
DeepSeek V3 671B671B473.8 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B473.8 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B868.8 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B164.2 GB4 tok/s4K ctx
moe
DeepSeek R1 671B671B473.8 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B207.5 GB4 tok/s4K ctx
moe
DeepSeek V2.5 236B236B207.5 GB4 tok/s4K ctx
moe
DeepSeek LLM 67B67B51.6 GB15 tok/s4K ctx
dense
Llama 4 Scout 17B 16E109B74.3 GB10 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B251.8 GB3 tok/s4K ctx
moe
Command R+ 104B104B71.8 GB5 tok/s4K ctx
dense
Command A 111B111B76.5 GB4 tok/s4K ctx
dense
Pixtral Large 124B124B85.9 GB3 tok/s4K ctx
dense
Qwen 2.5 VL 72B72B53.7 GB13 tok/s4K ctx
dense
Qwen 2.5 Math 72B72B53.7 GB13 tok/s4K ctx
dense
Nemotron 70B70B52.5 GB14 tok/s4K ctx
dense
Llama 3.1 405B405B259.6 GB2 tok/s4K ctx
dense
DeepSeek R1 Distill 70B70B52.5 GB14 tok/s4K ctx
dense
GPT-OSS 120B117B81.2 GB3 tok/s4K ctx
dense
DeepSeek V3.2671B414.7 GB2 tok/s4K ctx
moe
MiniMax M2.7
F0230B149.0 GB5 tok/s4K ctx
moe
Solar Open 100B
F0100B77.6 GB4 tok/s4K ctx
dense
K EXAONE 236B A23B
F0236B176.5 GB2 tok/s4K ctx
dense
Llama 3.3 70B Instruct
F070B55.8 GB11 tok/s4K ctx
dense
Baichuan M3 235B
F0235B175.8 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B175.8 GB2 tok/s4K ctx
dense
BaichuanMed OCR 72B i1
F072B57.3 GB11 tok/s4K ctx
dense
Solar Open 100B i1
F0100B77.6 GB4 tok/s4K ctx
dense
Solar Open 69B REAP i1
F069B55.1 GB12 tok/s4K ctx
dense
stabilityai japanese stablelm instruct beta 70b
F070B55.8 GB11 tok/s4K ctx
dense
Qwen3.5 122B A10B122B79.0 GB4 tok/s4K ctx
dense
Qwen3.5 397B A17B397B293.6 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs