AI Compatibility Tier List
AMD Instinct MI210 64GB
374 AI models ranked by how well they run on your 64 GB VRAM hardware. Best match: Qwen 3.6 35B A3B (score: 95).
374
Models tested
322
Can run
64 GB
VRAM
142
Best tok/s
Tier Distribution
S 23
A 61
B 52
C 186
F 52
S: 23A: 61B: 52C: 186F: 52
S
S Tier -- Perfect Fit (23 models)
These models run flawlessly with headroom to spare.
Qwen 3.6 35B A3B35B32.8 GB142 tok/s138K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B27.4 GB168 tok/s256K ctx
moe
Qwen 3.5 35B A3B35B30.1 GB154 tok/s131K ctx
moe
Qwen3-VL 30B A3B Instruct30B27.1 GB174 tok/s256K ctx
moe
Qwen3-Coder-Next80B57.6 GB75 tok/s86K ctx
moe
Qwen 3.5 27B27B26.9 GB73 tok/s131K ctx
dense
Qwen 3 32B32B30.7 GB62 tok/s131K ctx
dense
Qwen 3 30B A3B30.5B27.4 GB168 tok/s131K ctx
moe
Magistral Small 250724B24.4 GB82 tok/s131K ctx
dense
Devstral Small 2 24B Instruct24B24.4 GB82 tok/s256K ctx
dense
Qwen 2.5 VL 72B72B56.1 GB28 tok/s33K ctx
dense
Qwen 3.6 27B27B24.7 GB46 tok/s262K ctx
dense
Nemotron 3 Nano 30B30B28.0 GB65 tok/s131K ctx
dense
Gemma 4 31B30.7B40.7 GB39 tok/s41K ctx
dense
Devstral Small 1.124B24.4 GB82 tok/s131K ctx
dense
Nemotron Cascade 2 30B A3B30B28.5 GB172 tok/s210K ctx
moe
Qwen 3 14B14B18.3 GB141 tok/s131K ctx
dense
Qwen 3.5 9B9B15.0 GB126 tok/s131K ctx
dense
Phi-4-reasoning-plus 14B14.7B19.3 GB134 tok/s33K ctx
dense
GPT-OSS 20B21B22.6 GB214 tok/s128K ctx
moe
Qwen 3 8B8B14.4 GB112 tok/s131K ctx
dense
EXAONE 4.0 32B32B30.7 GB62 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B26.3 GB181 tok/s181K ctx
moe
A
A Tier -- Great Fit (61 models)
Excellent performance with comfortable memory margins.
Qwen 2.5 32B32B30.7 GB62 tok/s131K ctx
dense
Llama 3.3 70B70B54.9 GB28 tok/s46K ctx
dense
Qwen 3.5 4B4B11.9 GB56 tok/s131K ctx
dense
Kimi Linear 48B A3B48B37.5 GB38 tok/s473K ctx
linear
Gemma 3 27B27B35.0 GB44 tok/s57K ctx
dense
Mistral Small 3.2 24B24B24.4 GB82 tok/s131K ctx
vision
Codestral 2 25.0822B23.2 GB84 tok/s256K ctx
dense
OLMo 2 32B32B30.7 GB62 tok/s4K ctx
dense
Ministral 3 14B14B18.3 GB140 tok/s262K ctx
multimodal
Granite 4.1 30B30B29.5 GB65 tok/s131K ctx
dense
LFM2 24B24B24.4 GB82 tok/s131K ctx
dense
CogVLM2 19B19B21.3 GB103 tok/s8K ctx
dense
Nemotron Nano 8B8B14.1 GB112 tok/s131K ctx
dense
Mistral Small 24B24B24.4 GB82 tok/s33K ctx
dense
Llama 3.1 70B70B54.9 GB28 tok/s46K ctx
dense
Qwen 2.5 72B72B56.1 GB28 tok/s42K ctx
dense
Mistral Small 3.1 24B24B24.4 GB82 tok/s131K ctx
dense
Phi-4 Mini Reasoning 4B3.8B11.1 GB53 tok/s131K ctx
dense
Phi-4 14B14B18.9 GB140 tok/s16K ctx
dense
Granite Code 20B20B22.7 GB99 tok/s8K ctx
dense
Qwen 2.5 14B14B18.8 GB141 tok/s131K ctx
dense
InternVL2 8B8B14.1 GB112 tok/s8K ctx
dense
Qwen 2.5 Coder 32B32B30.7 GB62 tok/s131K ctx
dense
Gemma 3 12B12B19.5 GB99 tok/s131K ctx
dense
Granite Code 34B34B31.7 GB58 tok/s8K ctx
dense
StarCoder 15B15B32.7 GB105 tok/s8K ctx
dense
DeepSeek Coder V2 16B16B20.4 GB272 tok/s131K ctx
moe
Ministral 3 8B8B14.4 GB112 tok/s262K ctx
multimodal
MiniCPM-V 2.6 8B8B14.1 GB112 tok/s2K ctx
dense
Command R 35B35B31.1 GB57 tok/s131K ctx
dense
Nemotron Nano 9B v29B15.2 GB126 tok/s131K ctx
dense
DeepSeek R1 Distill 70B70B54.9 GB28 tok/s46K ctx
dense
DeepSeek R1 Distill 32B32B30.7 GB62 tok/s33K ctx
dense
CodeLlama 13B Instruct13B27.4 GB141 tok/s16K ctx
dense
SQLCoder 7B7B13.5 GB98 tok/s8K ctx
dense
MPT-30B-Instruct30B52.3 GB53 tok/s8K ctx
dense
Qwen 3 4B4B11.9 GB56 tok/s33K ctx
dense
Qwen 2.5 VL 7B7B12.4 GB98 tok/s33K ctx
dense
OLMo 2 13B13B17.7 GB152 tok/s33K ctx
dense
Jina Embeddings v30.57B10.4 GB8 tok/s8K ctx
dense
LLaVA 1.6 13B13B27.4 GB141 tok/s4K ctx
dense
Magistral 7B7B13.5 GB98 tok/s8K ctx
dense
CodeGeeX 4 9B9B13.4 GB126 tok/s131K ctx
dense
Gemma 4 E4B8B13.5 GB112 tok/s128K ctx
dense
BGE M30.57B9.6 GB8 tok/s8K ctx
dense
Falcon 40B Instruct40B37.9 GB43 tok/s8K ctx
dense
DeepSeek R1 Distill 14B14B18.8 GB141 tok/s33K ctx
dense
Nous Dolphin 13B13B28.9 GB121 tok/s16K ctx
dense
Qwen 2.5 7B7B12.4 GB98 tok/s131K ctx
dense
Nemotron 70B70B54.9 GB28 tok/s46K ctx
dense
Granite Code 8B8B14.1 GB112 tok/s8K ctx
dense
CodeLlama 7B Instruct7B19.4 GB98 tok/s16K ctx
dense
WizardLM 13B13B27.4 GB141 tok/s8K ctx
dense
Vicuna 13B13B27.4 GB141 tok/s4K ctx
dense
StarCoder 7B7B18.9 GB98 tok/s8K ctx
dense
DevStral 7B7B13.5 GB98 tok/s8K ctx
dense
Pixtral 12B12B17.1 GB164 tok/s131K ctx
dense
Codestral Mamba 7B7B12.1 GB98 tok/s262K ctx
state-space
Granite 4.1 8B8B14.6 GB112 tok/s131K ctx
dense
Nous Hermes 1.09B25.0 GB126 tok/s16K ctx
dense
Gemma 2 27B27B35.0 GB44 tok/s8K ctx
dense
B
B Tier -- Good Fit (52 models)
Solid performance, may need quantization or reduced context.
mxbai Embed Large0.34B9.5 GB5 tok/s1K ctx
dense
Mixtral 8x7B47B37.9 GB80 tok/s33K ctx
moe
Snowflake Arctic Embed L0.34B9.5 GB5 tok/s1K ctx
dense
InternLM 7B7B19.4 GB98 tok/s8K ctx
dense
InternLM Chat 7B7B19.4 GB98 tok/s8K ctx
dense
Qwen 2.5 Coder 3B3B11.3 GB42 tok/s131K ctx
dense
GLM-4 9B9B13.4 GB126 tok/s128K ctx
dense
Baichuan 13B13B28.9 GB121 tok/s8K ctx
dense
Llama 3.1 8B8B14.1 GB112 tok/s128K ctx
dense
Gemma 4 E2B5.1B10.9 GB71 tok/s128K ctx
dense
Solar 7B7B14.5 GB98 tok/s8K ctx
dense
BGE Large EN v1.50.34B9.5 GB5 tok/s1K ctx
dense
WizardMath 7B7B13.5 GB98 tok/s4K ctx
dense
OLMo 2 7B7B13.5 GB98 tok/s4K ctx
dense
Cerebras-GPT 13B13B26.4 GB121 tok/s78K ctx
dense
LLaVA 1.5 7B7B19.4 GB98 tok/s4K ctx
dense
Qwen 2.5 Coder 7B7B12.4 GB98 tok/s131K ctx
dense
Nomic Embed Text v1.50.14B8.1 GB2 tok/s8K ctx
dense
Gemma 3 4B4B11.8 GB56 tok/s128K ctx
dense
Ministral 3 3B3B9.9 GB42 tok/s262K ctx
multimodal
Phi 4 Mini 4B4B11.2 GB56 tok/s128K ctx
dense
MPT-7B-Instruct7B19.4 GB98 tok/s8K ctx
dense
Baichuan 7B7B19.4 GB98 tok/s8K ctx
dense
Qwen 3.5 2B2B10.2 GB28 tok/s131K ctx
dense
Yi 1.5 34B34B31.7 GB58 tok/s4K ctx
dense
Qwen 2.5 Math 72B72B56.1 GB28 tok/s4K ctx
dense
DeepSeek R1 Distill 8B8B14.1 GB112 tok/s33K ctx
dense
DeepSeek R1 Distill 7B7B12.4 GB98 tok/s33K ctx
dense
InternLM 20B20B40.0 GB79 tok/s8K ctx
dense
Samantha 7B7B13.5 GB98 tok/s4K ctx
dense
Qwen 2.5 Coder 14B14B18.8 GB141 tok/s131K ctx
dense
Phi 3 Mini 3.8B3.8B15.5 GB53 tok/s128K ctx
dense
Gemma 2 9B9B17.9 GB126 tok/s8K ctx
dense
Qwen 2.5 3B3B11.3 GB42 tok/s131K ctx
dense
Falcon 7B Instruct7B11.7 GB98 tok/s8K ctx
dense
Llama 3.2 11B Vision11B16.0 GB154 tok/s16K ctx
vision
Phi 3.5 Mini 4B4B15.6 GB56 tok/s128K ctx
dense
Llama 4 Scout 17B 16E109B76.7 GB22 tok/s4K ctx
moe
Qwen 3 1.7B1.7B10.0 GB24 tok/s33K ctx
dense
DeepSeek LLM 67B67B54.0 GB30 tok/s4K ctx
dense
All MiniLM L6 v20.02B7.6 GB2 tok/s0K ctx
dense
Mistral Nemo 12B12B17.1 GB164 tok/s128K ctx
dense
Granite 4.1 3B3B10.4 GB42 tok/s131K ctx
dense
Codestral 22B22B23.2 GB89 tok/s33K ctx
dense
Granite Code 3B3B11.6 GB42 tok/s8K ctx
dense
Phi 3 Medium 14B14B18.9 GB140 tok/s128K ctx
dense
Qwen 2.5 Coder 1.5B1.5B8.6 GB21 tok/s33K ctx
dense
Yi Coder 9B9B14.3 GB126 tok/s131K ctx
dense
Mistral 7B Instruct v0.37B13.5 GB98 tok/s8K ctx
dense
Ministral 8B8B14.4 GB112 tok/s131K ctx
dense
Aya Expanse 32B32B29.3 GB62 tok/s8K ctx
dense
Llama 3.2 3B3B10.8 GB42 tok/s128K ctx
dense
C
C Tier -- Tight Fit (186 models)
Runs with offloading or significant compromises.
Command R+ 104B104B74.2 GB11 tok/s4K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C5348B42.2 GB38 tok/s78K ctx
dense
Yi 34B Chat34B31.7 GB58 tok/s157K ctx
dense
Qwen3.5 35B A3B35B32.8 GB52 tok/s138K ctx
dense
Qwen3.5 35B A3B
C5135B32.8 GB52 tok/s138K ctx
dense
Granite 3.1 8B8B14.1 GB112 tok/s128K ctx
state-space
Yi 1.5 9B9B14.3 GB126 tok/s4K ctx
dense
StableLM 2 12B12B28.1 GB99 tok/s4K ctx
dense
Hermes 4.3 36B
C5136B33.5 GB51 tok/s132K ctx
dense
SmolLM3 3B3B11.1 GB42 tok/s128K ctx
dense
DeepSeek R1 1.5B1.5B8.6 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B8.3 GB15 tok/s4K ctx
dense
EXAONE 4.0 32B
C5032B30.6 GB57 tok/s159K ctx
dense
aya expanse 32b heretic MPOA i1
C5032B30.6 GB57 tok/s159K ctx
dense
Qwen3.5 27B27B26.9 GB68 tok/s203K ctx
dense
Baichuan M2 32B Q4 K M
C5032B30.6 GB57 tok/s159K ctx
dense
Qwen 2.5 Math 7B7B12.4 GB98 tok/s4K ctx
dense
baichuan inc Baichuan M2 32B
C5032B30.6 GB57 tok/s159K ctx
dense
gemma 3 27b it27B26.9 GB68 tok/s203K ctx
dense
gemma 3 27b it
C5027B26.9 GB68 tok/s203K ctx
dense
Llama 3.3 70B Instruct
C5070B58.2 GB26 tok/s27K ctx
dense
Mistral Small 3.2 24B Instruct 250624B24.8 GB76 tok/s239K ctx
dense
StarCoder2 15B15B19.3 GB115 tok/s16K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C5024B24.8 GB76 tok/s239K ctx
dense
Mistral Small 24B Instruct 2501
C5024B24.8 GB76 tok/s239K ctx
dense
Solar Open 69B REAP i1
C5069B57.5 GB27 tok/s29K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C5024B24.8 GB76 tok/s239K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C5024B24.8 GB76 tok/s239K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4924B24.8 GB76 tok/s239K ctx
dense
stabilityai japanese stablelm instruct beta 70b
C4970B58.2 GB26 tok/s27K ctx
dense
Codestral 22B v0.1
C4922B23.3 GB83 tok/s269K ctx
dense
Codestral 22B v0.1
C4922B23.3 GB83 tok/s269K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4924B24.8 GB76 tok/s239K ctx
dense
BaichuanMed OCR 72B i1
C4972B59.7 GB25 tok/s24K ctx
dense
Codestral 22B v0.1
C4922B23.3 GB83 tok/s269K ctx
dense
Codestral 22B v0.1 IMat
C4922B23.3 GB83 tok/s269K ctx
dense
Vicuna 7B7B19.4 GB98 tok/s4K ctx
dense
Codestral 22B v0.1 i1
C4922B23.3 GB83 tok/s269K ctx
dense
Codestral 21B Pruned i1
C4921B22.6 GB87 tok/s285K ctx
dense
internlm2 5 20b chat
C4920B21.8 GB91 tok/s304K ctx
dense
internlm2 limarp chat 20b
C4920B21.8 GB91 tok/s304K ctx
dense
internlm2 math plus 20b i1
C4920B21.8 GB91 tok/s304K ctx
dense
Codestral RAG 19B Pruned i1
C4819B21.1 GB96 tok/s324K ctx
dense
OpenChat 7B7B13.5 GB98 tok/s8K ctx
dense
Qwen 2.5 1.5B1.5B8.6 GB21 tok/s131K ctx
dense
Aya Expanse 8B8B14.1 GB112 tok/s8K ctx
dense
Phi 4 reasoning vision 15B
C4815B18.2 GB122 tok/s433K ctx
dense
StarCoder2 15B
C4815B18.2 GB122 tok/s433K ctx
dense
DeepSeek R1 Distill Qwen 14B14B17.5 GB130 tok/s470K ctx
dense
starcoder2 15b instruct v0.1
C4815B18.2 GB122 tok/s433K ctx
dense
starcoder2 15b i1
C4815B18.2 GB122 tok/s433K ctx
dense
NousResearch Hermes 4 14B
C4814B17.5 GB130 tok/s470K ctx
dense
starcoder2 15b instruct v0.1
C4815B18.2 GB122 tok/s433K ctx
dense
HelpingAI 15B i1
C4715B18.2 GB122 tok/s433K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4714B17.5 GB130 tok/s470K ctx
dense
internlm JanusCoder 14B
C4714B17.5 GB130 tok/s470K ctx
dense
gemma 3 12b it
C4712B16.0 GB152 tok/s562K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4714B17.5 GB130 tok/s470K ctx
dense
Dolphin 2.9 8B8B14.1 GB112 tok/s33K ctx
dense
OpenHermes 2.5 7B7B13.5 GB98 tok/s8K ctx
dense
Zephyr 7B Beta7B13.5 GB98 tok/s33K ctx
dense
Qwen3.5 9B9B13.8 GB126 tok/s777K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4710.7B15.1 GB150 tok/s640K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C479B13.8 GB126 tok/s777K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4710.7B15.1 GB150 tok/s640K ctx
dense
Qwen3.5 9B
C479B13.8 GB126 tok/s777K ctx
dense
Starling LM 7B7B13.5 GB98 tok/s8K ctx
dense
Gemma 3 1B1B8.3 GB14 tok/s33K ctx
dense
Meta Llama 3.1 8B Instruct
C478B13.1 GB112 tok/s884K ctx
dense
solar finalised finetuned Model 10.7B i1
C4710.7B15.1 GB150 tok/s640K ctx
dense
llava llama 3 8b v1 1
C478B13.1 GB112 tok/s884K ctx
dense
DeepSeek LLM 7B7B18.9 GB98 tok/s4K ctx
dense
Yi Coder 9B Chat
C479B13.8 GB126 tok/s777K ctx
dense
glm 4 9b chat 1m
C479B13.8 GB126 tok/s777K ctx
dense
SOLAR 10.7B v1.0
C4710.7B15.1 GB150 tok/s640K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B13.1 GB112 tok/s884K ctx
dense
Meta Llama 3 8B Instruct
C468B13.1 GB112 tok/s884K ctx
dense
DeepSeek R1 Distill Llama 8B8B13.1 GB112 tok/s884K ctx
dense
Dolphin3.0 Llama3.1 8B
C468B13.1 GB112 tok/s884K ctx
dense
Llama 3 8B Instruct 32k v0.1
C468B13.1 GB112 tok/s884K ctx
dense
Meta Llama 3.1 8B Instruct
C468B13.1 GB112 tok/s884K ctx
dense
vntl llama3 8b v2
C468B13.1 GB112 tok/s884K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C468B13.1 GB112 tok/s884K ctx
dense
Yi 1.5 9B Chat
C469B13.8 GB126 tok/s777K ctx
dense
Llama 2 7B Chat
C467B12.4 GB98 tok/s1023K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C468B13.1 GB112 tok/s884K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C468B13.1 GB112 tok/s884K ctx
dense
Mistral 7B Instruct v0.2
C467B12.4 GB98 tok/s1023K ctx
dense
Hermes 3 Llama 3.1 8B
C468B13.1 GB112 tok/s884K ctx
dense
Mistral 7B Instruct v0.3
C467B12.4 GB98 tok/s1023K ctx
dense
Helply 10.2b chat i1
C4610.2B14.7 GB143 tok/s676K ctx
dense
HelpingAI2.5 10B i1
C4610B14.6 GB140 tok/s691K ctx
dense
Hermes 2 Pro Llama 3 8B
C468B13.1 GB112 tok/s884K ctx
dense
Dolphin3.0 Llama3.1 8B
C468B13.1 GB112 tok/s884K ctx
dense
granite 8b code instruct 4k
C468B13.1 GB112 tok/s884K ctx
dense
dolphin 2.9.4 llama3.1 8b
C468B13.1 GB112 tok/s884K ctx
dense
Yi 9B Coder i1
C469B13.8 GB126 tok/s777K ctx
dense
Gemma 2 2B2B10.1 GB28 tok/s8K ctx
dense
openchat 3.6 8b 20240522 IMat
C468B13.1 GB112 tok/s884K ctx
dense
HelpingAI2 9B
C469B13.8 GB126 tok/s777K ctx
dense
aya expanse 8b
C468B13.1 GB112 tok/s884K ctx
dense
EXAONE 3.5 7.8B Instruct
C467.8B13.0 GB109 tok/s909K ctx
dense
Mistral 7B Instruct v0.3
C467B12.4 GB98 tok/s1023K ctx
dense
dolphin v2 8b abliterated i1
C468B13.1 GB112 tok/s884K ctx
dense
Hermes 2 Pro Mistral 7B
C467B12.4 GB98 tok/s1023K ctx
dense
HelpingAI2 9B i1
C469B13.8 GB126 tok/s777K ctx
dense
HelpingAI 9B 200k i1
C469B13.8 GB126 tok/s777K ctx
dense
EXAONE 3.5 7.8B Instruct
C467.8B13.0 GB109 tok/s909K ctx
dense
zephyr 7B beta
C467B12.4 GB98 tok/s1023K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C467.8B13.0 GB109 tok/s909K ctx
dense
HelpingAI 9B i1
C469B13.8 GB126 tok/s777K ctx
dense
aya expanse 8b orthogonal heretic i1
C468B13.1 GB112 tok/s884K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C467B12.4 GB98 tok/s1023K ctx
dense
exaone 3.0 7.8b it
C467.8B13.0 GB109 tok/s909K ctx
dense
Falcon H1R 7B
C467B12.4 GB98 tok/s1023K ctx
dense
falcon mamba 7b instruct Q4 K M
C467B12.4 GB98 tok/s1023K ctx
dense
aya expanse 8b orthogonal heretic
C468B13.1 GB112 tok/s884K ctx
dense
zephyr 7B alpha
C467B12.4 GB98 tok/s1023K ctx
dense
speechless zephyr code functionary 7b
C467B12.4 GB98 tok/s1023K ctx
dense
StarCoder2 7B
C467B12.4 GB98 tok/s1023K ctx
dense
japanese stablelm instruct gamma 7B
C467B12.4 GB98 tok/s1023K ctx
dense
internlm2 5 1 8b chat i1
C468B13.1 GB112 tok/s884K ctx
dense
Falcon H1 7B Instruct
C467B12.4 GB98 tok/s1023K ctx
dense
internlm3 8b instruct abliterated i1
C468B13.1 GB112 tok/s884K ctx
dense
Falcon H1R 7B7B12.4 GB98 tok/s1023K ctx
dense
Mamba Codestral 7B v0.1
C467B12.4 GB98 tok/s1023K ctx
dense
Neural Chat 7B7B13.5 GB98 tok/s8K ctx
dense
stabilityai japanese stablelm base gamma 7b
C467B12.4 GB98 tok/s1023K ctx
dense
baichuan2 7b chat
C467B12.4 GB98 tok/s1023K ctx
dense
DiscoPOP zephyr 7b gemma
C467B12.4 GB98 tok/s1023K ctx
dense
Yi 1.5 6B Chat
C466B11.7 GB84 tok/s1207K ctx
dense
starcoder2 7b
C467B12.4 GB98 tok/s1023K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C467B12.4 GB98 tok/s1023K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C467B12.4 GB98 tok/s1023K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C467B12.4 GB98 tok/s1023K ctx
dense
internlm2 math plus 7b IMat
C467B12.4 GB98 tok/s1023K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C467B12.4 GB98 tok/s1023K ctx
dense
internlm2 5 7b chat i1
C467B12.4 GB98 tok/s1023K ctx
dense
stablelm 2 zephyr 1 6b
C466B11.7 GB84 tok/s1207K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C467B12.4 GB98 tok/s1023K ctx
dense
MD Judge v0 2 internlm2 7b i1
C467B12.4 GB98 tok/s1023K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C457B12.4 GB98 tok/s1023K ctx
dense
jointpreferences mistral 7b sft helpful
C457B12.4 GB98 tok/s1023K ctx
dense
zephyr 7b dpo full i1
C457B12.4 GB98 tok/s1023K ctx
dense
blossom v3 baichuan2 7b i1
C457B12.4 GB98 tok/s1023K ctx
dense
blossom v1 baichuan 7b i1
C457B12.4 GB98 tok/s1023K ctx
dense
Yi 1.5 6B6B11.9 GB84 tok/s4K ctx
dense
Yi 1.5 6B Chat
C456B11.7 GB84 tok/s1207K ctx
dense
stablelm 2 1 6b chat imatrix
C456B11.7 GB84 tok/s1207K ctx
dense
Qwen 2.5 Coder 0.5B0.5B7.8 GB7 tok/s131K ctx
dense
StarCoder2 7B7B12.1 GB98 tok/s16K ctx
dense
HelpingAI2 6B
C456B11.7 GB84 tok/s1207K ctx
dense
HelpingAI2 6B i1
C456B11.7 GB84 tok/s1207K ctx
dense
Nemotron Mini 4B4B11.7 GB56 tok/s4K ctx
dense
HelpingAI2.5 5B i1
C445B10.9 GB70 tok/s1465K ctx
dense
Qwen3.5 4B4B10.2 GB56 tok/s1852K ctx
dense
gemma 3 4b it
C444B10.2 GB56 tok/s1852K ctx
dense
gemma 3 4b it
C444B10.2 GB56 tok/s1852K ctx
dense
Qwen 3 0.6B0.6B8.5 GB8 tok/s33K ctx
dense
Llama 3.2 1B1B8.4 GB14 tok/s128K ctx
dense
Llama 3.2 3B Instruct
C433B9.8 GB42 tok/s2482K ctx
dense
Qwen2.5 3B Instruct3B9.5 GB42 tok/s2497K ctx
dense
Llama 3.2 3B Instruct
C433B9.5 GB42 tok/s2497K ctx
dense
Ministral 3 3B Instruct 2512
C433B9.5 GB42 tok/s2497K ctx
dense
HELVETE 3B
C423B9.5 GB42 tok/s2497K ctx
dense
Hermes 3 Llama 3.2 3B
C423B9.5 GB42 tok/s2497K ctx
dense
stablelm zephyr 3b
C423B9.5 GB42 tok/s2497K ctx
dense
StarCoder2 3B
C423B9.5 GB42 tok/s2497K ctx
dense
AI21 Jamba Reasoning 3B
C423B9.5 GB42 tok/s2497K ctx
dense
stablelm 3b 4e1t
C423B9.5 GB42 tok/s2497K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C423B9.5 GB42 tok/s2497K ctx
dense
ai21labs AI21 Jamba2 3B
C423B9.5 GB42 tok/s2497K ctx
dense
HelpingAI 3B hindi i1
C423B9.5 GB42 tok/s2497K ctx
dense
AI21 Jamba2 3B
C423B9.5 GB42 tok/s2497K ctx
dense
HelpingAI 3B hindi
C423B9.5 GB42 tok/s2497K ctx
dense
AI21 Jamba2 3B i1
C423B9.5 GB42 tok/s2497K ctx
dense
gemma 2 2b it
C412B9.2 GB28 tok/s3759K ctx
dense
EXAONE 3.5 2.4B Instruct
C412.4B9.0 GB34 tok/s3142K ctx
dense
gemma 2b2B8.8 GB28 tok/s3787K ctx
dense
Gemmasutra Mini 2B v1
C412B8.8 GB28 tok/s3787K ctx
dense
StarCoder2 3B3B9.6 GB42 tok/s16K ctx
dense
gemma 2 2b it
C412B8.8 GB28 tok/s3787K ctx
dense
Qwen 3.5 0.6B0.6B8.5 GB8 tok/s131K ctx
dense
Qwen2.5 1.5B Instruct1.5B8.4 GB21 tok/s5078K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B8.4 GB21 tok/s5078K ctx
dense
Yi Coder 1.5B Chat
C401.5B8.4 GB21 tok/s5078K ctx
dense
F
F Tier -- Can't Run (52 models)
Too heavy for this hardware.
stablelm 2 zephyr 1.6b
D401.6B8.5 GB22 tok/s4755K ctx
dense
logos16v2 stablelm2 1.6b i1
D401.6B8.5 GB22 tok/s4755K ctx
dense
Yi Coder 1.5B
D401.5B8.4 GB21 tok/s5078K ctx
dense
Falcon H1 1.5B Instruct1.5B8.4 GB21 tok/s5078K ctx
dense
TinyLlama 1.1B Chat v1.0
D391.1B8.1 GB15 tok/s6954K ctx
dense
Qwen 2.5 0.5B0.5B7.8 GB7 tok/s131K ctx
dense
EXAONE 4.0 1.2B
D391.2B8.2 GB17 tok/s6368K ctx
dense
Llama 3.2 1B Instruct Q8 0
D391B8.2 GB14 tok/s7629K ctx
dense
TinyLlama 1.1B Chat v0.3
D391.1B8.1 GB15 tok/s6954K ctx
dense
TinyLlama 1.1B Chat v0.61.1B8.1 GB15 tok/s6954K ctx
dense
Llama 3.2 1B Instruct
D391B8.0 GB14 tok/s7658K ctx
dense
gemma 3 1b it
D391B8.0 GB14 tok/s7658K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D381.1B8.1 GB15 tok/s6954K ctx
dense
Falcon3 1B Instruct abliterated
D381B8.0 GB14 tok/s7658K ctx
dense
SmolVLM 500M Instruct
D370.5B7.8 GB7 tok/s9006K ctx
dense
embeddinggemma 300M
D350.3B7.6 GB4 tok/s9033K ctx
dense
granite embedding 107m multilingual
D330.11B7.5 GB2 tok/s9062K ctx
dense
Falcon H1 Tiny 90M Instruct
D330.09B7.5 GB2 tok/s9063K ctx
dense
Qwen 3 235B A22B235B153.5 GB3 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B303.0 GB2 tok/s4K ctx
moe
Qwen 3.5 122B A10B122B84.2 GB19 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B252.3 GB2 tok/s4K ctx
moe
Mixtral 8x22B141B96.7 GB9 tok/s4K ctx
moe
Devstral 2 123B Instruct123B87.7 GB6 tok/s4K ctx
dense
Mistral Small 4 119B119B85.3 GB20 tok/s4K ctx
moe
Leanstral 119B A6B119B88.7 GB17 tok/s4K ctx
moe
Mistral Large 3675B424.4 GB2 tok/s4K ctx
moe
GLM-5744B480.2 GB2 tok/s4K ctx
moe
GLM-5.1754B486.3 GB2 tok/s4K ctx
moe
Kimi K2.51000B624.7 GB2 tok/s4K ctx
moe
Kimi K2.61000B624.7 GB2 tok/s4K ctx
moe
DeepSeek V3 671B671B476.2 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B476.2 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B871.2 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B166.6 GB4 tok/s4K ctx
moe
DeepSeek R1 671B671B476.2 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B209.9 GB3 tok/s4K ctx
moe
DeepSeek V2.5 236B236B209.9 GB3 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B254.2 GB2 tok/s4K ctx
moe
Command A 111B111B78.9 GB9 tok/s4K ctx
dense
Pixtral Large 124B124B88.3 GB6 tok/s4K ctx
dense
Llama 3.1 405B405B262.0 GB2 tok/s4K ctx
dense
GPT-OSS 120B117B83.6 GB7 tok/s4K ctx
dense
DeepSeek V3.2671B417.1 GB2 tok/s4K ctx
moe
MiniMax M2.7
F0230B151.4 GB4 tok/s4K ctx
moe
Solar Open 100B
F0100B80.0 GB9 tok/s4K ctx
dense
K EXAONE 236B A23B
F0236B178.9 GB2 tok/s4K ctx
dense
Baichuan M3 235B
F0235B178.2 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B178.2 GB2 tok/s4K ctx
dense
Solar Open 100B i1
F0100B80.0 GB9 tok/s4K ctx
dense
Qwen3.5 122B A10B122B81.4 GB8 tok/s4K ctx
dense
Qwen3.5 397B A17B397B296.0 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs