AI Compatibility Tier List
AMD Instinct MI325X 256GB
374 AI models ranked by how well they run on your 256 GB VRAM hardware. Best match: DeepSeek V4 Flash (score: 99).
374
Models tested
341
Can run
256 GB
VRAM
94
Best tok/s
Tier Distribution
S 28
A 55
B 66
C 192
F 33
S: 28A: 55B: 66C: 192F: 33
S
S Tier -- Perfect Fit (28 models)
These models run flawlessly with headroom to spare.
DeepSeek V4 Flash284B185.8 GB94 tok/s872K ctx
moe
Qwen 3 235B A22B235B172.7 GB89 tok/s131K ctx
moe
Qwen 3.5 122B A10B122B103.4 GB176 tok/s131K ctx
moe
Devstral 2 123B Instruct123B106.9 GB64 tok/s256K ctx
dense
MiniMax M2.7
S93230B170.6 GB101 tok/s205K ctx
moe
Mistral Small 4 119B119B104.5 GB191 tok/s256K ctx
moe
GPT-OSS 120B117B102.8 GB67 tok/s131K ctx
dense
Command A 111B111B98.1 GB71 tok/s262K ctx
dense
DeepSeek Coder V2 236B236B229.1 GB82 tok/s23K ctx
moe
Pixtral Large 124B124B107.5 GB63 tok/s131K ctx
dense
Qwen3-Coder 30B A3B Instruct30.5B46.6 GB662 tok/s256K ctx
moe
Qwen 2.5 VL 72B72B75.3 GB109 tok/s33K ctx
dense
Qwen 3.6 35B A3B35B52.0 GB557 tok/s262K ctx
moe
Qwen 3.5 27B27B46.1 GB287 tok/s131K ctx
dense
Qwen3-Coder-Next80B76.8 GB296 tok/s256K ctx
moe
Qwen 3.6 27B27B43.9 GB179 tok/s262K ctx
dense
Qwen3-VL 30B A3B Instruct30B46.3 GB685 tok/s256K ctx
moe
DeepSeek V2.5 236B236B229.1 GB82 tok/s23K ctx
moe
Qwen 3.5 35B A3B35B49.3 GB605 tok/s131K ctx
moe
Qwen 3 32B32B49.9 GB244 tok/s131K ctx
dense
Magistral Small 250724B43.6 GB322 tok/s131K ctx
dense
Devstral Small 2 24B Instruct24B43.6 GB322 tok/s256K ctx
dense
Leanstral 119B A6B119B107.9 GB176 tok/s256K ctx
moe
Qwen 3.5 9B9B34.2 GB126 tok/s131K ctx
dense
Qwen 3 30B A3B30.5B46.6 GB662 tok/s131K ctx
moe
Nemotron 3 Nano 30B30B47.2 GB257 tok/s131K ctx
dense
Qwen 3 14B14B37.5 GB196 tok/s131K ctx
dense
Devstral Small 1.124B43.6 GB322 tok/s131K ctx
dense
A
A Tier -- Great Fit (55 models)
Excellent performance with comfortable memory margins.
Phi-4-reasoning-plus 14B14.7B38.5 GB206 tok/s33K ctx
dense
Qwen 3 8B8B33.6 GB112 tok/s131K ctx
dense
Qwen 3.5 397B A17B397B271.5 GB39 tok/s4K ctx
moe
GPT-OSS 20B21B41.8 GB841 tok/s128K ctx
moe
Gemma 4 31B30.7B59.9 GB153 tok/s230K ctx
dense
Nemotron Cascade 2 30B A3B30B47.7 GB677 tok/s262K ctx
moe
Qwen 3.5 4B4B31.1 GB56 tok/s131K ctx
dense
Llama 3.3 70B70B74.1 GB112 tok/s128K ctx
dense
EXAONE 4.0 32B32B49.9 GB242 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B45.5 GB711 tok/s256K ctx
moe
Ministral 3 14B14B37.5 GB196 tok/s262K ctx
multimodal
Phi-4 Mini Reasoning 4B3.8B30.3 GB53 tok/s131K ctx
dense
Nemotron Nano 8B8B33.3 GB112 tok/s131K ctx
dense
Codestral 2 25.0822B42.4 GB308 tok/s256K ctx
dense
Qwen 2.5 32B32B49.9 GB242 tok/s131K ctx
dense
Mistral Small 3.2 24B24B43.6 GB322 tok/s131K ctx
vision
Qwen 2.5 72B72B75.3 GB109 tok/s131K ctx
dense
Llama 3.1 70B70B74.1 GB112 tok/s128K ctx
dense
Kimi Linear 48B A3B48B56.7 GB150 tok/s1049K ctx
linear
Gemma 3 27B27B54.2 GB174 tok/s131K ctx
dense
LFM2 24B24B43.6 GB322 tok/s131K ctx
dense
CogVLM2 19B19B40.5 GB266 tok/s8K ctx
dense
OLMo 2 32B32B49.9 GB242 tok/s4K ctx
dense
Granite 4.1 30B30B48.7 GB257 tok/s131K ctx
dense
Llama 4 Scout 17B 16E109B95.9 GB168 tok/s890K ctx
moe
Mistral Small 24B24B43.6 GB322 tok/s33K ctx
dense
InternVL2 8B8B33.3 GB112 tok/s8K ctx
dense
Mistral Small 3.1 24B24B43.6 GB322 tok/s131K ctx
dense
Phi-4 14B14B38.1 GB196 tok/s16K ctx
dense
Qwen 2.5 14B14B38.0 GB196 tok/s131K ctx
dense
Llama 4 Maverick 17B 128E400B273.4 GB38 tok/s4K ctx
moe
Ministral 3 8B8B33.6 GB112 tok/s262K ctx
multimodal
MiniCPM-V 2.6 8B8B33.3 GB112 tok/s2K ctx
dense
Granite Code 20B20B41.9 GB280 tok/s8K ctx
dense
Gemma 3 12B12B38.7 GB168 tok/s131K ctx
dense
DeepSeek R1 Distill 70B70B74.1 GB112 tok/s131K ctx
dense
Qwen 3 4B4B31.1 GB56 tok/s33K ctx
dense
SQLCoder 7B7B32.7 GB98 tok/s8K ctx
dense
Nemotron Nano 9B v29B34.4 GB126 tok/s131K ctx
dense
Qwen 2.5 VL 7B7B31.6 GB98 tok/s33K ctx
dense
Qwen 2.5 Coder 32B32B49.9 GB242 tok/s131K ctx
dense
DeepSeek Coder V2 16B16B39.6 GB1069 tok/s131K ctx
moe
Jina Embeddings v30.57B29.6 GB8 tok/s8K ctx
dense
Granite Code 34B34B50.9 GB229 tok/s8K ctx
dense
BGE M30.57B28.8 GB8 tok/s8K ctx
dense
Magistral 7B7B32.7 GB98 tok/s8K ctx
dense
CodeGeeX 4 9B9B32.6 GB126 tok/s131K ctx
dense
Command R 35B35B50.3 GB223 tok/s131K ctx
dense
Gemma 4 E4B8B32.7 GB112 tok/s128K ctx
dense
OLMo 2 13B13B36.9 GB182 tok/s33K ctx
dense
DeepSeek R1 Distill 32B32B49.9 GB242 tok/s33K ctx
dense
StarCoder 15B15B51.9 GB210 tok/s8K ctx
dense
Llama 3.1 405B405B281.2 GB12 tok/s4K ctx
dense
CodeLlama 13B Instruct13B46.6 GB182 tok/s16K ctx
dense
Qwen 2.5 7B7B31.6 GB98 tok/s131K ctx
dense
B
B Tier -- Good Fit (66 models)
Solid performance, may need quantization or reduced context.
Granite Code 8B8B33.3 GB112 tok/s8K ctx
dense
Nemotron 70B70B74.1 GB112 tok/s131K ctx
dense
DevStral 7B7B32.7 GB98 tok/s8K ctx
dense
Codestral Mamba 7B7B31.3 GB98 tok/s262K ctx
state-space
DeepSeek R1 Distill 14B14B38.0 GB196 tok/s33K ctx
dense
mxbai Embed Large0.34B28.7 GB5 tok/s1K ctx
dense
LLaVA 1.6 13B13B46.6 GB182 tok/s4K ctx
dense
Granite 4.1 8B8B33.8 GB112 tok/s131K ctx
dense
Pixtral 12B12B36.3 GB168 tok/s131K ctx
dense
MPT-30B-Instruct30B71.5 GB207 tok/s8K ctx
dense
CodeLlama 7B Instruct7B38.6 GB98 tok/s16K ctx
dense
StarCoder 7B7B38.1 GB98 tok/s8K ctx
dense
Snowflake Arctic Embed L0.34B28.7 GB5 tok/s1K ctx
dense
Qwen 2.5 Coder 3B3B30.5 GB42 tok/s131K ctx
dense
Falcon 40B Instruct40B57.1 GB169 tok/s8K ctx
dense
Nous Dolphin 13B13B48.1 GB182 tok/s16K ctx
dense
Gemma 4 E2B5.1B30.1 GB71 tok/s128K ctx
dense
BGE Large EN v1.50.34B28.7 GB5 tok/s1K ctx
dense
WizardLM 13B13B46.6 GB182 tok/s8K ctx
dense
Vicuna 13B13B46.6 GB182 tok/s4K ctx
dense
Command R+ 104B104B93.4 GB75 tok/s131K ctx
dense
Nous Hermes 1.09B44.2 GB126 tok/s16K ctx
dense
Gemma 2 27B27B54.2 GB174 tok/s8K ctx
dense
GLM-4 9B9B32.6 GB126 tok/s128K ctx
dense
InternLM 7B7B38.6 GB98 tok/s8K ctx
dense
InternLM Chat 7B7B38.6 GB98 tok/s8K ctx
dense
Llama 3.1 8B8B33.3 GB112 tok/s128K ctx
dense
Solar 7B7B33.7 GB98 tok/s8K ctx
dense
WizardMath 7B7B32.7 GB98 tok/s4K ctx
dense
OLMo 2 7B7B32.7 GB98 tok/s4K ctx
dense
Nomic Embed Text v1.50.14B27.3 GB2 tok/s8K ctx
dense
Mixtral 8x22B141B115.9 GB106 tok/s66K ctx
moe
Qwen 2.5 Coder 7B7B31.6 GB98 tok/s131K ctx
dense
Ministral 3 3B3B29.1 GB42 tok/s262K ctx
multimodal
Gemma 3 4B4B31.0 GB56 tok/s128K ctx
dense
Phi 4 Mini 4B4B30.4 GB56 tok/s128K ctx
dense
LLaVA 1.5 7B7B38.6 GB98 tok/s4K ctx
dense
Qwen 3.5 2B2B29.4 GB28 tok/s131K ctx
dense
Mixtral 8x7B47B57.1 GB315 tok/s33K ctx
moe
Baichuan 13B13B48.1 GB182 tok/s8K ctx
dense
Cerebras-GPT 13B13B45.6 GB182 tok/s131K ctx
dense
Qwen 2.5 Math 72B72B75.3 GB109 tok/s4K ctx
dense
DeepSeek R1 Distill 7B7B31.6 GB98 tok/s33K ctx
dense
MPT-7B-Instruct7B38.6 GB98 tok/s8K ctx
dense
Baichuan 7B7B38.6 GB98 tok/s8K ctx
dense
DeepSeek R1 Distill 8B8B33.3 GB112 tok/s33K ctx
dense
Samantha 7B7B32.7 GB98 tok/s4K ctx
dense
Qwen 2.5 3B3B30.5 GB42 tok/s131K ctx
dense
Falcon 7B Instruct7B30.9 GB98 tok/s8K ctx
dense
Phi 3 Mini 3.8B3.8B34.7 GB53 tok/s128K ctx
dense
Qwen 3 1.7B1.7B29.2 GB24 tok/s33K ctx
dense
All MiniLM L6 v20.02B26.8 GB2 tok/s0K ctx
dense
Gemma 2 9B9B37.1 GB126 tok/s8K ctx
dense
Qwen 2.5 Coder 14B14B38.0 GB196 tok/s131K ctx
dense
Llama 3.2 11B Vision11B35.2 GB154 tok/s16K ctx
vision
Phi 3.5 Mini 4B4B34.8 GB56 tok/s128K ctx
dense
Granite 4.1 3B3B29.6 GB42 tok/s131K ctx
dense
Yi 1.5 34B34B50.9 GB229 tok/s4K ctx
dense
Qwen 2.5 Coder 1.5B1.5B27.8 GB21 tok/s33K ctx
dense
Granite Code 3B3B30.8 GB42 tok/s8K ctx
dense
DeepSeek LLM 67B67B73.2 GB117 tok/s4K ctx
dense
Mistral Nemo 12B12B36.3 GB168 tok/s128K ctx
dense
Yi Coder 9B9B33.5 GB126 tok/s131K ctx
dense
Mistral 7B Instruct v0.37B32.7 GB98 tok/s8K ctx
dense
Phi 3 Medium 14B14B38.1 GB196 tok/s128K ctx
dense
Codestral 22B22B42.4 GB308 tok/s33K ctx
dense
C
C Tier -- Tight Fit (192 models)
Runs with offloading or significant compromises.
InternLM 20B20B59.2 GB280 tok/s8K ctx
dense
Llama 3.2 3B3B30.0 GB42 tok/s128K ctx
dense
Ministral 8B8B33.6 GB112 tok/s131K ctx
dense
K EXAONE 236B A23B
C53236B198.1 GB30 tok/s49K ctx
dense
Baichuan M3 235B
C53235B197.4 GB31 tok/s50K ctx
dense
Baichuan M3 235B i1
C53235B197.4 GB31 tok/s50K ctx
dense
Aya Expanse 32B32B48.5 GB244 tok/s8K ctx
dense
DeepSeek R1 1.5B1.5B27.8 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B27.5 GB15 tok/s4K ctx
dense
Qwen3.5 122B A10B122B100.6 GB68 tok/s190K ctx
dense
SmolLM3 3B3B30.3 GB42 tok/s128K ctx
dense
Granite 3.1 8B8B33.3 GB112 tok/s128K ctx
state-space
Solar Open 100B
C49100B99.2 GB72 tok/s230K ctx
dense
Solar Open 100B i1
C49100B99.2 GB72 tok/s230K ctx
dense
Yi 1.5 9B9B33.5 GB126 tok/s4K ctx
dense
Qwen 2.5 Math 7B7B31.6 GB98 tok/s4K ctx
dense
Llama 3.3 70B Instruct
C4870B77.4 GB103 tok/s364K ctx
dense
Solar Open 69B REAP i1
C4869B76.7 GB104 tok/s371K ctx
dense
stabilityai japanese stablelm instruct beta 70b
C4870B77.4 GB103 tok/s364K ctx
dense
BaichuanMed OCR 72B i1
C4872B78.9 GB100 tok/s352K ctx
dense
Yi 34B Chat34B50.9 GB229 tok/s200K ctx
dense
Qwen 2.5 1.5B1.5B27.8 GB21 tok/s131K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C4748B61.4 GB150 tok/s570K ctx
dense
OpenChat 7B7B32.7 GB98 tok/s8K ctx
dense
Qwen3.5 35B A3B35B52.0 GB205 tok/s812K ctx
dense
StarCoder2 15B15B38.5 GB210 tok/s16K ctx
dense
Qwen3.5 35B A3B
C4635B52.0 GB205 tok/s812K ctx
dense
Aya Expanse 8B8B33.3 GB112 tok/s8K ctx
dense
Gemma 3 1B1B27.5 GB14 tok/s33K ctx
dense
Qwen3.5 27B27B46.1 GB266 tok/s1077K ctx
dense
Hermes 4.3 36B
C4636B52.7 GB200 tok/s787K ctx
dense
gemma 3 27b it27B46.1 GB266 tok/s1077K ctx
dense
EXAONE 4.0 32B
C4632B49.8 GB224 tok/s896K ctx
dense
gemma 3 27b it
C4627B46.1 GB266 tok/s1077K ctx
dense
aya expanse 32b heretic MPOA i1
C4632B49.8 GB224 tok/s896K ctx
dense
Baichuan M2 32B Q4 K M
C4632B49.8 GB224 tok/s896K ctx
dense
Mistral Small 3.2 24B Instruct 250624B44.0 GB299 tok/s1222K ctx
dense
Vicuna 7B7B38.6 GB98 tok/s4K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4624B44.0 GB299 tok/s1222K ctx
dense
baichuan inc Baichuan M2 32B
C4632B49.8 GB224 tok/s896K ctx
dense
Mistral Small 24B Instruct 2501
C4624B44.0 GB299 tok/s1222K ctx
dense
StableLM 2 12B12B47.3 GB168 tok/s4K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4624B44.0 GB299 tok/s1222K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4624B44.0 GB299 tok/s1222K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4524B44.0 GB299 tok/s1222K ctx
dense
OpenHermes 2.5 7B7B32.7 GB98 tok/s8K ctx
dense
Codestral 22B v0.1
C4522B42.5 GB308 tok/s1341K ctx
dense
Codestral 22B v0.1
C4522B42.5 GB308 tok/s1341K ctx
dense
Dolphin 2.9 8B8B33.3 GB112 tok/s33K ctx
dense
Gemma 2 2B2B29.3 GB28 tok/s8K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4524B44.0 GB299 tok/s1222K ctx
dense
Codestral 22B v0.1
C4522B42.5 GB308 tok/s1341K ctx
dense
Zephyr 7B Beta7B32.7 GB98 tok/s33K ctx
dense
DeepSeek R1 Distill Qwen 14B14B36.7 GB196 tok/s2155K ctx
dense
Codestral 22B v0.1 IMat
C4522B42.5 GB308 tok/s1341K ctx
dense
Qwen3.5 9B9B33.0 GB126 tok/s3398K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C459B33.0 GB126 tok/s3398K ctx
dense
Qwen 2.5 Coder 0.5B0.5B27.0 GB7 tok/s131K ctx
dense
gemma 3 12b it
C4512B35.2 GB168 tok/s2528K ctx
dense
Meta Llama 3.1 8B Instruct
C458B32.3 GB112 tok/s3834K ctx
dense
Phi 4 reasoning vision 15B
C4515B37.4 GB210 tok/s2006K ctx
dense
internlm2 5 20b chat
C4520B41.0 GB280 tok/s1483K ctx
dense
Starling LM 7B7B32.7 GB98 tok/s8K ctx
dense
Codestral 22B v0.1 i1
C4522B42.5 GB308 tok/s1341K ctx
dense
StarCoder2 15B
C4515B37.4 GB210 tok/s2006K ctx
dense
Codestral 21B Pruned i1
C4521B41.8 GB294 tok/s1409K ctx
dense
llava llama 3 8b v1 1
C458B32.3 GB112 tok/s3834K ctx
dense
internlm2 limarp chat 20b
C4520B41.0 GB280 tok/s1483K ctx
dense
Qwen3.5 9B
C459B33.0 GB126 tok/s3398K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B32.3 GB112 tok/s3834K ctx
dense
NousResearch Hermes 4 14B
C4514B36.7 GB196 tok/s2155K ctx
dense
Llama 2 7B Chat
C457B31.6 GB98 tok/s4393K ctx
dense
internlm2 math plus 20b i1
C4520B41.0 GB280 tok/s1483K ctx
dense
Meta Llama 3 8B Instruct
C458B32.3 GB112 tok/s3834K ctx
dense
Mistral 7B Instruct v0.2
C457B31.6 GB98 tok/s4393K ctx
dense
DeepSeek R1 Distill Llama 8B8B32.3 GB112 tok/s3834K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B32.3 GB112 tok/s3834K ctx
dense
Llama 3 8B Instruct 32k v0.1
C458B32.3 GB112 tok/s3834K ctx
dense
Mistral 7B Instruct v0.3
C457B31.6 GB98 tok/s4393K ctx
dense
Meta Llama 3.1 8B Instruct
C458B32.3 GB112 tok/s3834K ctx
dense
vntl llama3 8b v2
C458B32.3 GB112 tok/s3834K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B32.3 GB112 tok/s3834K ctx
dense
Codestral RAG 19B Pruned i1
C4519B40.3 GB266 tok/s1566K ctx
dense
Yi Coder 9B Chat
C459B33.0 GB126 tok/s3398K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4510.7B34.3 GB150 tok/s2845K ctx
dense
starcoder2 15b instruct v0.1
C4515B37.4 GB210 tok/s2006K ctx
dense
glm 4 9b chat 1m
C459B33.0 GB126 tok/s3398K ctx
dense
starcoder2 15b i1
C4515B37.4 GB210 tok/s2006K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4510.7B34.3 GB150 tok/s2845K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B32.3 GB112 tok/s3834K ctx
dense
starcoder2 15b instruct v0.1
C4515B37.4 GB210 tok/s2006K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C458B32.3 GB112 tok/s3834K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4514B36.7 GB196 tok/s2155K ctx
dense
Hermes 3 Llama 3.1 8B
C458B32.3 GB112 tok/s3834K ctx
dense
internlm JanusCoder 14B
C4514B36.7 GB196 tok/s2155K ctx
dense
HelpingAI 15B i1
C4515B37.4 GB210 tok/s2006K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4514B36.7 GB196 tok/s2155K ctx
dense
Yi 1.5 9B Chat
C459B33.0 GB126 tok/s3398K ctx
dense
Hermes 2 Pro Llama 3 8B
C458B32.3 GB112 tok/s3834K ctx
dense
Mistral 7B Instruct v0.3
C457B31.6 GB98 tok/s4393K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B32.3 GB112 tok/s3834K ctx
dense
Hermes 2 Pro Mistral 7B
C457B31.6 GB98 tok/s4393K ctx
dense
granite 8b code instruct 4k
C448B32.3 GB112 tok/s3834K ctx
dense
zephyr 7B beta
C447B31.6 GB98 tok/s4393K ctx
dense
dolphin 2.9.4 llama3.1 8b
C448B32.3 GB112 tok/s3834K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C447B31.6 GB98 tok/s4393K ctx
dense
Falcon H1R 7B
C447B31.6 GB98 tok/s4393K ctx
dense
solar finalised finetuned Model 10.7B i1
C4410.7B34.3 GB150 tok/s2845K ctx
dense
openchat 3.6 8b 20240522 IMat
C448B32.3 GB112 tok/s3834K ctx
dense
falcon mamba 7b instruct Q4 K M
C447B31.6 GB98 tok/s4393K ctx
dense
EXAONE 3.5 7.8B Instruct
C447.8B32.2 GB109 tok/s3934K ctx
dense
aya expanse 8b
C448B32.3 GB112 tok/s3834K ctx
dense
zephyr 7B alpha
C447B31.6 GB98 tok/s4393K ctx
dense
speechless zephyr code functionary 7b
C447B31.6 GB98 tok/s4393K ctx
dense
SOLAR 10.7B v1.0
C4410.7B34.3 GB150 tok/s2845K ctx
dense
dolphin v2 8b abliterated i1
C448B32.3 GB112 tok/s3834K ctx
dense
EXAONE 3.5 7.8B Instruct
C447.8B32.2 GB109 tok/s3934K ctx
dense
StarCoder2 7B
C447B31.6 GB98 tok/s4393K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C447.8B32.2 GB109 tok/s3934K ctx
dense
japanese stablelm instruct gamma 7B
C447B31.6 GB98 tok/s4393K ctx
dense
Yi 9B Coder i1
C449B33.0 GB126 tok/s3398K ctx
dense
aya expanse 8b orthogonal heretic i1
C448B32.3 GB112 tok/s3834K ctx
dense
exaone 3.0 7.8b it
C447.8B32.2 GB109 tok/s3934K ctx
dense
Yi 1.5 6B Chat
C446B30.9 GB84 tok/s5139K ctx
dense
Falcon H1 7B Instruct
C447B31.6 GB98 tok/s4393K ctx
dense
Falcon H1R 7B7B31.6 GB98 tok/s4393K ctx
dense
HelpingAI2 9B
C449B33.0 GB126 tok/s3398K ctx
dense
stablelm 2 zephyr 1 6b
C446B30.9 GB84 tok/s5139K ctx
dense
aya expanse 8b orthogonal heretic
C448B32.3 GB112 tok/s3834K ctx
dense
HelpingAI2.5 10B i1
C4410B33.8 GB140 tok/s3050K ctx
dense
Helply 10.2b chat i1
C4410.2B33.9 GB143 tok/s2989K ctx
dense
HelpingAI2 9B i1
C449B33.0 GB126 tok/s3398K ctx
dense
Mamba Codestral 7B v0.1
C447B31.6 GB98 tok/s4393K ctx
dense
stabilityai japanese stablelm base gamma 7b
C447B31.6 GB98 tok/s4393K ctx
dense
HelpingAI 9B 200k i1
C449B33.0 GB126 tok/s3398K ctx
dense
baichuan2 7b chat
C447B31.6 GB98 tok/s4393K ctx
dense
DiscoPOP zephyr 7b gemma
C447B31.6 GB98 tok/s4393K ctx
dense
HelpingAI 9B i1
C449B33.0 GB126 tok/s3398K ctx
dense
internlm2 5 1 8b chat i1
C448B32.3 GB112 tok/s3834K ctx
dense
internlm3 8b instruct abliterated i1
C448B32.3 GB112 tok/s3834K ctx
dense
starcoder2 7b
C447B31.6 GB98 tok/s4393K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C447B31.6 GB98 tok/s4393K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C447B31.6 GB98 tok/s4393K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C447B31.6 GB98 tok/s4393K ctx
dense
internlm2 math plus 7b IMat
C447B31.6 GB98 tok/s4393K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C447B31.6 GB98 tok/s4393K ctx
dense
Yi 1.5 6B Chat
C446B30.9 GB84 tok/s5139K ctx
dense
internlm2 5 7b chat i1
C447B31.6 GB98 tok/s4393K ctx
dense
Yi 1.5 6B6B31.1 GB84 tok/s4K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C447B31.6 GB98 tok/s4393K ctx
dense
stablelm 2 1 6b chat imatrix
C446B30.9 GB84 tok/s5139K ctx
dense
MD Judge v0 2 internlm2 7b i1
C447B31.6 GB98 tok/s4393K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C447B31.6 GB98 tok/s4393K ctx
dense
jointpreferences mistral 7b sft helpful
C447B31.6 GB98 tok/s4393K ctx
dense
zephyr 7b dpo full i1
C447B31.6 GB98 tok/s4393K ctx
dense
blossom v3 baichuan2 7b i1
C447B31.6 GB98 tok/s4393K ctx
dense
blossom v1 baichuan 7b i1
C447B31.6 GB98 tok/s4393K ctx
dense
Neural Chat 7B7B32.7 GB98 tok/s8K ctx
dense
HelpingAI2 6B
C446B30.9 GB84 tok/s5139K ctx
dense
StarCoder2 7B7B31.3 GB98 tok/s16K ctx
dense
HelpingAI2 6B i1
C446B30.9 GB84 tok/s5139K ctx
dense
Nemotron Mini 4B4B30.9 GB56 tok/s4K ctx
dense
DeepSeek LLM 7B7B38.1 GB98 tok/s4K ctx
dense
Qwen3.5 4B4B29.4 GB56 tok/s7750K ctx
dense
HelpingAI2.5 5B i1
C435B30.1 GB70 tok/s6184K ctx
dense
gemma 3 4b it
C434B29.4 GB56 tok/s7750K ctx
dense
gemma 3 4b it
C434B29.4 GB56 tok/s7750K ctx
dense
Qwen 3 0.6B0.6B27.7 GB8 tok/s33K ctx
dense
Llama 3.2 1B1B27.6 GB14 tok/s128K ctx
dense
Llama 3.2 3B Instruct
C423B29.0 GB42 tok/s10346K ctx
dense
Qwen2.5 3B Instruct3B28.7 GB42 tok/s10362K ctx
dense
Llama 3.2 3B Instruct
C423B28.7 GB42 tok/s10362K ctx
dense
Ministral 3 3B Instruct 2512
C423B28.7 GB42 tok/s10362K ctx
dense
HELVETE 3B
C423B28.7 GB42 tok/s10362K ctx
dense
Hermes 3 Llama 3.2 3B
C423B28.7 GB42 tok/s10362K ctx
dense
stablelm zephyr 3b
C423B28.7 GB42 tok/s10362K ctx
dense
StarCoder2 3B
C423B28.7 GB42 tok/s10362K ctx
dense
AI21 Jamba Reasoning 3B
C413B28.7 GB42 tok/s10362K ctx
dense
stablelm 3b 4e1t
C413B28.7 GB42 tok/s10362K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C413B28.7 GB42 tok/s10362K ctx
dense
ai21labs AI21 Jamba2 3B
C413B28.7 GB42 tok/s10362K ctx
dense
HelpingAI 3B hindi i1
C413B28.7 GB42 tok/s10362K ctx
dense
AI21 Jamba2 3B
C413B28.7 GB42 tok/s10362K ctx
dense
HelpingAI 3B hindi
C413B28.7 GB42 tok/s10362K ctx
dense
AI21 Jamba2 3B i1
C413B28.7 GB42 tok/s10362K ctx
dense
EXAONE 3.5 2.4B Instruct
C412.4B28.2 GB34 tok/s12973K ctx
dense
gemma 2 2b it
C412B28.4 GB28 tok/s15555K ctx
dense
gemma 2b2B28.0 GB28 tok/s15584K ctx
dense
Gemmasutra Mini 2B v1
C412B28.0 GB28 tok/s15584K ctx
dense
Qwen 3.5 0.6B0.6B27.7 GB8 tok/s131K ctx
dense
gemma 2 2b it
C402B28.0 GB28 tok/s15584K ctx
dense
StarCoder2 3B3B28.8 GB42 tok/s16K ctx
dense
F
F Tier -- Can't Run (33 models)
Too heavy for this hardware.
Qwen2.5 1.5B Instruct1.5B27.6 GB21 tok/s20806K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B27.6 GB21 tok/s20806K ctx
dense
Yi Coder 1.5B Chat
D401.5B27.6 GB21 tok/s20806K ctx
dense
stablelm 2 zephyr 1.6b
D391.6B27.7 GB22 tok/s19501K ctx
dense
logos16v2 stablelm2 1.6b i1
D391.6B27.7 GB22 tok/s19501K ctx
dense
Yi Coder 1.5B
D391.5B27.6 GB21 tok/s20806K ctx
dense
Falcon H1 1.5B Instruct1.5B27.6 GB21 tok/s20806K ctx
dense
TinyLlama 1.1B Chat v1.0
D391.1B27.3 GB15 tok/s28403K ctx
dense
Qwen 2.5 0.5B0.5B27.0 GB7 tok/s131K ctx
dense
EXAONE 4.0 1.2B
D391.2B27.4 GB17 tok/s26029K ctx
dense
Llama 3.2 1B Instruct Q8 0
D381B27.4 GB14 tok/s31222K ctx
dense
TinyLlama 1.1B Chat v0.3
D381.1B27.3 GB15 tok/s28403K ctx
dense
TinyLlama 1.1B Chat v0.61.1B27.3 GB15 tok/s28403K ctx
dense
Llama 3.2 1B Instruct
D381B27.2 GB14 tok/s31251K ctx
dense
gemma 3 1b it
D381B27.2 GB14 tok/s31251K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D381.1B27.3 GB15 tok/s28403K ctx
dense
Falcon3 1B Instruct abliterated
D381B27.2 GB14 tok/s31251K ctx
dense
SmolVLM 500M Instruct
D360.5B27.0 GB7 tok/s36654K ctx
dense
embeddinggemma 300M
D350.3B26.8 GB4 tok/s36681K ctx
dense
granite embedding 107m multilingual
D330.11B26.7 GB2 tok/s36710K ctx
dense
Falcon H1 Tiny 90M Instruct
D320.09B26.7 GB2 tok/s36711K ctx
dense
Qwen3-Coder 480B A35B Instruct480B322.2 GB21 tok/s4K ctx
moe
Mistral Large 3675B443.6 GB8 tok/s4K ctx
moe
GLM-5744B499.4 GB5 tok/s4K ctx
moe
GLM-5.1754B505.5 GB5 tok/s4K ctx
moe
Kimi K2.51000B643.9 GB4 tok/s4K ctx
moe
Kimi K2.61000B643.9 GB4 tok/s4K ctx
moe
DeepSeek V3 671B671B495.4 GB6 tok/s4K ctx
moe
DeepSeek V3.1 671B671B495.4 GB6 tok/s4K ctx
moe
DeepSeek V4 Pro1600B890.4 GB3 tok/s4K ctx
moe
DeepSeek R1 671B671B495.4 GB6 tok/s4K ctx
moe
DeepSeek V3.2671B436.3 GB8 tok/s4K ctx
moe
Qwen3.5 397B A17B397B315.2 GB9 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs