AI Compatibility Tier List
RTX 3070 8GB
374 AI models ranked by how well they run on your 8 GB VRAM hardware. Best match: Qwen 3.5 4B (score: 95).
374
Models tested
195
Can run
8 GB
VRAM
48
Best tok/s
Tier Distribution
S 3
A 31
B 27
C 134
F 179
S: 3A: 31B: 27C: 134F: 179
S
S Tier -- Perfect Fit (3 models)
These models run flawlessly with headroom to spare.
A
A Tier -- Great Fit (31 models)
Excellent performance with comfortable memory margins.
Jina Embeddings v30.57B4.8 GB7 tok/s8K ctx
dense
Qwen 2.5 VL 7B7B6.8 GB67 tok/s33K ctx
dense
SQLCoder 7B7B7.9 GB66 tok/s8K ctx
dense
Codestral Mamba 7B7B6.5 GB84 tok/s67K ctx
state-space
Qwen 3.5 9B9B9.4 GB28 tok/s6K ctx
dense
Magistral 7B7B7.9 GB66 tok/s8K ctx
dense
BGE M30.57B4.0 GB7 tok/s8K ctx
dense
Qwen 3 8B8B8.8 GB31 tok/s10K ctx
dense
CodeGeeX 4 9B9B7.8 GB53 tok/s21K ctx
dense
Qwen 2.5 Coder 3B3B5.7 GB36 tok/s33K ctx
dense
Gemma 4 E4B8B7.9 GB44 tok/s18K ctx
dense
Qwen 2.5 7B7B6.8 GB67 tok/s38K ctx
dense
DevStral 7B7B7.9 GB66 tok/s8K ctx
dense
Gemma 4 E2B5.1B5.3 GB61 tok/s96K ctx
dense
mxbai Embed Large0.34B3.9 GB4 tok/s1K ctx
dense
Nemotron Nano 8B8B8.5 GB38 tok/s12K ctx
dense
Gemma 3 4B4B6.2 GB48 tok/s30K ctx
dense
Snowflake Arctic Embed L0.34B3.9 GB4 tok/s1K ctx
dense
Phi 4 Mini 4B4B5.6 GB48 tok/s42K ctx
dense
BGE Large EN v1.50.34B3.9 GB4 tok/s1K ctx
dense
GLM-4 9B9B7.8 GB53 tok/s21K ctx
dense
WizardMath 7B7B7.9 GB66 tok/s4K ctx
dense
OLMo 2 7B7B7.9 GB66 tok/s4K ctx
dense
InternVL2 8B8B8.5 GB38 tok/s8K ctx
dense
Ministral 3 3B3B4.3 GB36 tok/s98K ctx
multimodal
Qwen 2.5 Coder 7B7B6.8 GB67 tok/s38K ctx
dense
Qwen 2.5 3B3B5.7 GB36 tok/s33K ctx
dense
Qwen 3.5 2B2B4.6 GB24 tok/s48K ctx
dense
Falcon 7B Instruct7B6.1 GB68 tok/s8K ctx
dense
MiniCPM-V 2.6 8B8B8.5 GB38 tok/s2K ctx
dense
Ministral 3 8B8B8.8 GB36 tok/s10K ctx
multimodal
B
B Tier -- Good Fit (27 models)
Solid performance, may need quantization or reduced context.
Granite Code 3B3B6.0 GB36 tok/s8K ctx
dense
DeepSeek R1 Distill 7B7B6.8 GB67 tok/s33K ctx
dense
Nomic Embed Text v1.50.14B2.5 GB2 tok/s8K ctx
dense
Samantha 7B7B7.9 GB66 tok/s4K ctx
dense
Granite 4.1 3B3B4.8 GB36 tok/s59K ctx
dense
Qwen 3 1.7B1.7B4.4 GB20 tok/s33K ctx
dense
Granite Code 8B8B8.5 GB38 tok/s8K ctx
dense
Llama 3.2 3B3B5.2 GB36 tok/s42K ctx
dense
Mistral 7B Instruct v0.37B7.9 GB66 tok/s8K ctx
dense
Granite 4.1 8B8B9.0 GB34 tok/s9K ctx
dense
Qwen 2.5 Coder 1.5B1.5B3.0 GB18 tok/s33K ctx
dense
All MiniLM L6 v20.02B2.0 GB2 tok/s0K ctx
dense
Llama 3.1 8B8B8.5 GB38 tok/s12K ctx
dense
Solar 7B7B8.9 GB40 tok/s8K ctx
dense
SmolLM3 3B3B5.5 GB36 tok/s37K ctx
dense
DeepSeek R1 Distill 8B8B8.5 GB38 tok/s12K ctx
dense
Qwen 2.5 Math 7B7B6.8 GB67 tok/s4K ctx
dense
Yi 1.5 6B Chat
B566B6.1 GB72 tok/s60K ctx
dense
stablelm 2 zephyr 1 6b
B566B6.1 GB72 tok/s60K ctx
dense
Yi 1.5 6B Chat
B566B6.1 GB72 tok/s60K ctx
dense
stablelm 2 1 6b chat imatrix
B566B6.1 GB72 tok/s60K ctx
dense
Yi 1.5 6B6B6.3 GB72 tok/s4K ctx
dense
HelpingAI2 6B
B566B6.1 GB72 tok/s60K ctx
dense
HelpingAI2 6B i1
B566B6.1 GB72 tok/s60K ctx
dense
Nemotron Mini 4B4B6.1 GB48 tok/s4K ctx
dense
DeepSeek R1 1.5B1.5B3.0 GB18 tok/s33K ctx
dense
StarCoder2 7B7B6.5 GB67 tok/s16K ctx
dense
C
C Tier -- Tight Fit (134 models)
Runs with offloading or significant compromises.
OpenChat 7B7B7.9 GB66 tok/s8K ctx
dense
TinyLlama 1.1B1.1B2.7 GB13 tok/s4K ctx
dense
Gemma 2 2B2B4.5 GB24 tok/s8K ctx
dense
HelpingAI2.5 5B i1
C545B5.3 GB60 tok/s89K ctx
dense
OpenHermes 2.5 7B7B7.9 GB66 tok/s8K ctx
dense
falcon mamba 7b instruct Q4 K M
C537B6.8 GB84 tok/s40K ctx
dense
Zephyr 7B Beta7B7.9 GB66 tok/s17K ctx
dense
Mamba Codestral 7B v0.1
C537B6.8 GB84 tok/s40K ctx
dense
Starling LM 7B7B7.9 GB66 tok/s8K ctx
dense
Llama 2 7B Chat
C537B6.8 GB62 tok/s40K ctx
dense
Mistral 7B Instruct v0.2
C537B6.8 GB62 tok/s40K ctx
dense
Qwen 2.5 1.5B1.5B3.0 GB18 tok/s131K ctx
dense
Mistral 7B Instruct v0.3
C537B6.8 GB62 tok/s40K ctx
dense
Yi Coder 9B9B8.7 GB33 tok/s9K ctx
dense
Mistral 7B Instruct v0.3
C537B6.8 GB62 tok/s40K ctx
dense
Hermes 2 Pro Mistral 7B
C537B6.8 GB62 tok/s40K ctx
dense
zephyr 7B beta
C537B6.8 GB62 tok/s40K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C537B6.8 GB62 tok/s40K ctx
dense
Meta Llama 3.1 8B Instruct
C538B7.5 GB54 tok/s24K ctx
dense
Falcon H1R 7B
C527B6.8 GB62 tok/s40K ctx
dense
llava llama 3 8b v1 1
C528B7.5 GB54 tok/s24K ctx
dense
zephyr 7B alpha
C527B6.8 GB62 tok/s40K ctx
dense
speechless zephyr code functionary 7b
C527B6.8 GB62 tok/s40K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B7.5 GB54 tok/s24K ctx
dense
Qwen3.5 4B4B4.6 GB48 tok/s132K ctx
dense
StarCoder2 7B
C527B6.8 GB62 tok/s40K ctx
dense
Meta Llama 3 8B Instruct
C528B7.5 GB54 tok/s24K ctx
dense
japanese stablelm instruct gamma 7B
C527B6.8 GB62 tok/s40K ctx
dense
DeepSeek R1 Distill Llama 8B8B7.5 GB54 tok/s24K ctx
dense
Dolphin3.0 Llama3.1 8B
C528B7.5 GB54 tok/s24K ctx
dense
Falcon H1 7B Instruct
C527B6.8 GB62 tok/s40K ctx
dense
Llama 3 8B Instruct 32k v0.1
C528B7.5 GB54 tok/s24K ctx
dense
Falcon H1R 7B7B6.8 GB62 tok/s40K ctx
dense
Meta Llama 3.1 8B Instruct
C528B7.5 GB54 tok/s24K ctx
dense
vntl llama3 8b v2
C528B7.5 GB54 tok/s24K ctx
dense
gemma 3 4b it
C524B4.6 GB48 tok/s132K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C528B7.5 GB54 tok/s24K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C528B7.5 GB54 tok/s24K ctx
dense
stabilityai japanese stablelm base gamma 7b
C527B6.8 GB62 tok/s40K ctx
dense
gemma 3 4b it
C524B4.6 GB48 tok/s132K ctx
dense
baichuan2 7b chat
C527B6.8 GB62 tok/s40K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C528B7.5 GB54 tok/s24K ctx
dense
DiscoPOP zephyr 7b gemma
C527B6.8 GB62 tok/s40K ctx
dense
Hermes 3 Llama 3.1 8B
C528B7.5 GB54 tok/s24K ctx
dense
starcoder2 7b
C527B6.8 GB62 tok/s40K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C527B6.8 GB62 tok/s40K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C527B6.8 GB62 tok/s40K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C527B6.8 GB62 tok/s40K ctx
dense
internlm2 math plus 7b IMat
C527B6.8 GB62 tok/s40K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C527B6.8 GB62 tok/s40K ctx
dense
Hermes 2 Pro Llama 3 8B
C528B7.5 GB54 tok/s24K ctx
dense
internlm2 5 7b chat i1
C527B6.8 GB62 tok/s40K ctx
dense
Dolphin3.0 Llama3.1 8B
C528B7.5 GB54 tok/s24K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C527B6.8 GB62 tok/s40K ctx
dense
Neural Chat 7B7B7.9 GB66 tok/s8K ctx
dense
MD Judge v0 2 internlm2 7b i1
C527B6.8 GB62 tok/s40K ctx
dense
granite 8b code instruct 4k
C528B7.5 GB54 tok/s24K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C527B6.8 GB62 tok/s40K ctx
dense
jointpreferences mistral 7b sft helpful
C527B6.8 GB62 tok/s40K ctx
dense
zephyr 7b dpo full i1
C527B6.8 GB62 tok/s40K ctx
dense
dolphin 2.9.4 llama3.1 8b
C528B7.5 GB54 tok/s24K ctx
dense
blossom v3 baichuan2 7b i1
C527B6.8 GB62 tok/s40K ctx
dense
EXAONE 3.5 7.8B Instruct
C527.8B7.4 GB55 tok/s27K ctx
dense
blossom v1 baichuan 7b i1
C527B6.8 GB62 tok/s40K ctx
dense
EXAONE 3.5 7.8B Instruct
C527.8B7.4 GB55 tok/s27K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C527.8B7.4 GB55 tok/s27K ctx
dense
openchat 3.6 8b 20240522 IMat
C528B7.5 GB54 tok/s24K ctx
dense
aya expanse 8b
C528B7.5 GB54 tok/s24K ctx
dense
exaone 3.0 7.8b it
C527.8B7.4 GB55 tok/s27K ctx
dense
dolphin v2 8b abliterated i1
C528B7.5 GB54 tok/s24K ctx
dense
aya expanse 8b orthogonal heretic i1
C528B7.5 GB54 tok/s24K ctx
dense
aya expanse 8b orthogonal heretic
C528B7.5 GB54 tok/s24K ctx
dense
internlm2 5 1 8b chat i1
C528B7.5 GB54 tok/s24K ctx
dense
internlm3 8b instruct abliterated i1
C528B7.5 GB54 tok/s24K ctx
dense
Qwen3.5 9B9B8.2 GB34 tok/s12K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C519B8.2 GB34 tok/s12K ctx
dense
Qwen3.5 9B
C519B8.2 GB34 tok/s12K ctx
dense
Yi Coder 9B Chat
C519B8.2 GB34 tok/s12K ctx
dense
glm 4 9b chat 1m
C519B8.2 GB34 tok/s12K ctx
dense
Gemma 3 1B1B2.7 GB12 tok/s33K ctx
dense
Yi 1.5 9B Chat
C519B8.2 GB34 tok/s12K ctx
dense
Llama 3.2 3B Instruct
C503B4.2 GB36 tok/s188K ctx
dense
Ministral 8B8B8.8 GB36 tok/s10K ctx
dense
Yi 9B Coder i1
C509B8.2 GB34 tok/s12K ctx
dense
HelpingAI2 9B
C509B8.2 GB34 tok/s12K ctx
dense
HelpingAI2 9B i1
C509B8.2 GB34 tok/s12K ctx
dense
HelpingAI 9B 200k i1
C509B8.2 GB34 tok/s12K ctx
dense
HelpingAI 9B i1
C509B8.2 GB34 tok/s12K ctx
dense
Qwen2.5 3B Instruct3B3.9 GB36 tok/s203K ctx
dense
Llama 3.2 3B Instruct
C493B3.9 GB36 tok/s203K ctx
dense
Ministral 3 3B Instruct 2512
C493B3.9 GB36 tok/s203K ctx
dense
HELVETE 3B
C493B3.9 GB36 tok/s203K ctx
dense
Hermes 3 Llama 3.2 3B
C493B3.9 GB36 tok/s203K ctx
dense
stablelm zephyr 3b
C493B3.9 GB36 tok/s203K ctx
dense
StarCoder2 3B
C493B3.9 GB36 tok/s203K ctx
dense
AI21 Jamba Reasoning 3B
C493B3.9 GB36 tok/s203K ctx
dense
stablelm 3b 4e1t
C493B3.9 GB36 tok/s203K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C493B3.9 GB36 tok/s203K ctx
dense
ai21labs AI21 Jamba2 3B
C493B3.9 GB36 tok/s203K ctx
dense
HelpingAI 3B hindi i1
C493B3.9 GB36 tok/s203K ctx
dense
AI21 Jamba2 3B
C493B3.9 GB36 tok/s203K ctx
dense
HelpingAI 3B hindi
C493B3.9 GB36 tok/s203K ctx
dense
AI21 Jamba2 3B i1
C493B3.9 GB36 tok/s203K ctx
dense
Qwen 2.5 Coder 0.5B0.5B2.2 GB6 tok/s131K ctx
dense
StarCoder2 3B3B4.0 GB36 tok/s16K ctx
dense
Qwen 3 0.6B0.6B2.9 GB7 tok/s33K ctx
dense
gemma 2 2b it
C482B3.6 GB24 tok/s318K ctx
dense
Llama 3.2 1B1B2.8 GB12 tok/s128K ctx
dense
EXAONE 3.5 2.4B Instruct
C472.4B3.4 GB29 tok/s275K ctx
dense
Granite 3.1 8B8B8.5 GB52 tok/s12K ctx
state-space
gemma 2b2B3.2 GB24 tok/s347K ctx
dense
Gemmasutra Mini 2B v1
C462B3.2 GB24 tok/s347K ctx
dense
gemma 2 2b it
C462B3.2 GB24 tok/s347K ctx
dense
Qwen 3.5 0.6B0.6B2.9 GB7 tok/s111K ctx
dense
Yi 1.5 9B9B8.7 GB33 tok/s4K ctx
dense
Qwen2.5 1.5B Instruct1.5B2.8 GB18 tok/s490K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B2.8 GB18 tok/s490K ctx
dense
Yi Coder 1.5B Chat
C441.5B2.8 GB18 tok/s490K ctx
dense
stablelm 2 zephyr 1.6b
C441.6B2.9 GB19 tok/s454K ctx
dense
logos16v2 stablelm2 1.6b i1
C441.6B2.9 GB19 tok/s454K ctx
dense
Yi Coder 1.5B
C441.5B2.8 GB18 tok/s490K ctx
dense
Falcon H1 1.5B Instruct1.5B2.8 GB18 tok/s490K ctx
dense
EXAONE 4.0 1.2B
C431.2B2.6 GB14 tok/s634K ctx
dense
Llama 3.2 1B Instruct Q8 0
C431B2.6 GB12 tok/s748K ctx
dense
TinyLlama 1.1B Chat v1.0
C431.1B2.5 GB13 tok/s699K ctx
dense
Aya Expanse 8B8B8.5 GB38 tok/s8K ctx
dense
TinyLlama 1.1B Chat v0.3
C421.1B2.5 GB13 tok/s699K ctx
dense
TinyLlama 1.1B Chat v0.61.1B2.5 GB13 tok/s699K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
C421.1B2.5 GB13 tok/s699K ctx
dense
Llama 3.2 1B Instruct
C421B2.4 GB12 tok/s777K ctx
dense
gemma 3 1b it
C421B2.4 GB12 tok/s777K ctx
dense
Qwen 2.5 0.5B0.5B2.2 GB6 tok/s131K ctx
dense
Dolphin 2.9 8B8B8.5 GB38 tok/s12K ctx
dense
Falcon3 1B Instruct abliterated
C421B2.4 GB12 tok/s777K ctx
dense
F
F Tier -- Can't Run (179 models)
Too heavy for this hardware.
Helply 10.2b chat i1
D4010.2B9.1 GB28 tok/s4K ctx
dense
SmolVLM 500M Instruct
D390.5B2.2 GB6 tok/s942K ctx
dense
HelpingAI2.5 10B i1
D3910B9.0 GB25 tok/s4K ctx
dense
embeddinggemma 300M
D370.3B2.0 GB4 tok/s969K ctx
dense
granite embedding 107m multilingual
D350.11B1.9 GB2 tok/s998K ctx
dense
Falcon H1 Tiny 90M Instruct
D350.09B1.9 GB2 tok/s999K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
D3410.7B9.5 GB25 tok/s4K ctx
dense
Nous Hermes 2 SOLAR 10.7B
D3410.7B9.5 GB25 tok/s4K ctx
dense
solar finalised finetuned Model 10.7B i1
D3310.7B9.5 GB25 tok/s4K ctx
dense
SOLAR 10.7B v1.0
D3310.7B9.5 GB25 tok/s4K ctx
dense
Llama 3.1 70B70B49.3 GB2 tok/s4K ctx
dense
Llama 3.3 70B70B49.3 GB2 tok/s4K ctx
dense
Llama 3.2 11B Vision11B10.4 GB21 tok/s4K ctx
vision
Qwen 2.5 14B14B13.2 GB10 tok/s4K ctx
dense
Qwen 2.5 32B32B25.1 GB2 tok/s4K ctx
dense
Qwen 2.5 72B72B50.5 GB2 tok/s4K ctx
dense
Qwen 2.5 Coder 14B14B13.2 GB10 tok/s4K ctx
dense
Qwen 2.5 Coder 32B32B25.1 GB2 tok/s4K ctx
dense
Qwen 3 14B14B12.7 GB11 tok/s4K ctx
dense
Qwen 3 32B32B25.1 GB2 tok/s4K ctx
dense
Qwen 3 235B A22B235B147.9 GB2 tok/s4K ctx
moe
Qwen 3 30B A3B30.5B21.8 GB5 tok/s4K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B21.8 GB5 tok/s4K ctx
moe
Qwen3-Coder-Next80B52.0 GB3 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B297.4 GB2 tok/s4K ctx
moe
Qwen3-VL 30B A3B Instruct30B21.5 GB7 tok/s4K ctx
moe
Qwen 3.5 27B27B21.3 GB2 tok/s4K ctx
dense
Qwen 3.5 35B A3B35B24.5 GB7 tok/s4K ctx
moe
Qwen 3.6 35B A3B35B27.2 GB6 tok/s4K ctx
moe
Qwen 3.6 27B27B19.1 GB2 tok/s4K ctx
dense
Qwen 3.5 122B A10B122B78.6 GB2 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B246.7 GB2 tok/s4K ctx
moe
Gemma 2 9B9B12.3 GB12 tok/s4K ctx
dense
Gemma 2 27B27B29.4 GB2 tok/s4K ctx
dense
Gemma 3 12B12B13.9 GB8 tok/s4K ctx
dense
Gemma 3 27B27B29.4 GB2 tok/s4K ctx
dense
Ministral 3 14B14B12.7 GB11 tok/s4K ctx
multimodal
Mistral Nemo 12B12B11.5 GB16 tok/s4K ctx
dense
Mistral Small 24B24B18.8 GB3 tok/s4K ctx
dense
Mistral Small 3.2 24B24B18.8 GB3 tok/s4K ctx
vision
Mixtral 8x7B47B32.3 GB3 tok/s4K ctx
moe
Mixtral 8x22B141B91.1 GB2 tok/s4K ctx
moe
DeepSeek R1 Distill 14B14B13.2 GB10 tok/s4K ctx
dense
DeepSeek R1 Distill 32B32B25.1 GB2 tok/s4K ctx
dense
Phi 3.5 Mini 4B4B10.0 GB48 tok/s11K ctx
dense
Phi-4 14B14B13.3 GB10 tok/s4K ctx
dense
Phi-4-reasoning-plus 14B14.7B13.7 GB9 tok/s4K ctx
dense
Granite 4.1 30B30B23.9 GB2 tok/s4K ctx
dense
OLMo 2 13B13B12.1 GB13 tok/s4K ctx
dense
Command R 35B35B25.5 GB2 tok/s4K ctx
dense
Falcon 40B Instruct40B32.3 GB2 tok/s4K ctx
dense
WizardLM 13B13B21.8 GB6 tok/s4K ctx
dense
StarCoder 7B7B13.3 GB16 tok/s4K ctx
dense
StarCoder 15B15B27.1 GB4 tok/s4K ctx
dense
CodeLlama 7B Instruct7B13.8 GB15 tok/s4K ctx
dense
CodeLlama 13B Instruct13B21.8 GB6 tok/s4K ctx
dense
StarCoder2 15B15B13.7 GB7 tok/s4K ctx
dense
MPT-7B-Instruct7B13.8 GB15 tok/s4K ctx
dense
MPT-30B-Instruct30B46.7 GB2 tok/s4K ctx
dense
InternLM 7B7B13.8 GB15 tok/s4K ctx
dense
InternLM Chat 7B7B13.8 GB15 tok/s4K ctx
dense
Devstral Small 1.124B18.8 GB3 tok/s4K ctx
dense
Devstral Small 2 24B Instruct24B18.8 GB3 tok/s4K ctx
dense
Devstral 2 123B Instruct123B82.1 GB2 tok/s4K ctx
dense
Mistral Small 4 119B119B79.7 GB2 tok/s4K ctx
moe
Leanstral 119B A6B119B83.1 GB2 tok/s4K ctx
moe
Mistral Large 3675B418.8 GB2 tok/s4K ctx
moe
Magistral Small 250724B18.8 GB3 tok/s4K ctx
dense
Codestral 2 25.0822B17.6 GB3 tok/s4K ctx
dense
GLM-5744B474.6 GB2 tok/s4K ctx
moe
GLM-5.1754B480.7 GB2 tok/s4K ctx
moe
Kimi K2.51000B619.1 GB2 tok/s4K ctx
moe
Kimi K2.61000B619.1 GB2 tok/s4K ctx
moe
Kimi Linear 48B A3B48B31.9 GB2 tok/s4K ctx
linear
Nous Hermes 1.09B19.4 GB7 tok/s4K ctx
dense
Nous Dolphin 13B13B23.3 GB5 tok/s4K ctx
dense
Baichuan 7B7B13.8 GB15 tok/s4K ctx
dense
Baichuan 13B13B23.3 GB5 tok/s4K ctx
dense
Cerebras-GPT 13B13B20.8 GB5 tok/s4K ctx
dense
StableLM 2 12B12B22.5 GB5 tok/s4K ctx
dense
InternLM 20B20B34.4 GB3 tok/s4K ctx
dense
Yi 34B Chat34B26.1 GB2 tok/s4K ctx
dense
DeepSeek V3 671B671B470.6 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B470.6 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B865.6 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B161.0 GB2 tok/s4K ctx
moe
DeepSeek R1 671B671B470.6 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 16B16B14.8 GB14 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B204.3 GB2 tok/s4K ctx
moe
DeepSeek V2.5 236B236B204.3 GB2 tok/s4K ctx
moe
DeepSeek LLM 7B7B13.3 GB16 tok/s4K ctx
dense
DeepSeek LLM 67B67B48.4 GB2 tok/s4K ctx
dense
Llama 4 Scout 17B 16E109B71.1 GB2 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B248.6 GB2 tok/s4K ctx
moe
Command R+ 104B104B68.6 GB2 tok/s4K ctx
dense
Command A 111B111B73.3 GB2 tok/s4K ctx
dense
Phi 3 Mini 3.8B3.8B9.9 GB46 tok/s11K ctx
dense
Phi 3 Medium 14B14B13.3 GB10 tok/s4K ctx
dense
Yi 1.5 34B34B26.1 GB2 tok/s4K ctx
dense
Codestral 22B22B17.6 GB3 tok/s4K ctx
dense
Pixtral 12B12B11.5 GB16 tok/s4K ctx
dense
Pixtral Large 124B124B82.7 GB2 tok/s4K ctx
dense
Qwen 2.5 VL 72B72B50.5 GB2 tok/s4K ctx
dense
Qwen 2.5 Math 72B72B50.5 GB2 tok/s4K ctx
dense
Nemotron 70B70B49.3 GB2 tok/s4K ctx
dense
LLaVA 1.5 7B7B13.8 GB15 tok/s4K ctx
dense
LLaVA 1.6 13B13B21.8 GB6 tok/s4K ctx
dense
CogVLM2 19B19B15.7 GB5 tok/s4K ctx
dense
Vicuna 7B7B13.8 GB15 tok/s4K ctx
dense
Vicuna 13B13B21.8 GB6 tok/s4K ctx
dense
Granite Code 20B20B17.1 GB4 tok/s4K ctx
dense
GPT-OSS 20B21B17.0 GB8 tok/s4K ctx
moe
OLMo 2 32B32B25.1 GB2 tok/s4K ctx
dense
Nemotron Nano 9B v29B9.6 GB26 tok/s5K ctx
dense
Nemotron 3 Nano 30B30B22.4 GB2 tok/s4K ctx
dense
EXAONE 4.0 32B32B25.1 GB2 tok/s4K ctx
dense
LFM2 24B24B18.8 GB3 tok/s4K ctx
dense
Llama 3.1 405B405B256.4 GB2 tok/s4K ctx
dense
DeepSeek R1 Distill 70B70B49.3 GB2 tok/s4K ctx
dense
Granite Code 34B34B26.1 GB2 tok/s4K ctx
dense
GPT-OSS 120B117B78.0 GB2 tok/s4K ctx
dense
Aya Expanse 32B32B23.7 GB2 tok/s4K ctx
dense
Mistral Small 3.1 24B24B18.8 GB3 tok/s4K ctx
dense
Gemma 4 26B A4B25.2B20.7 GB7 tok/s4K ctx
moe
Gemma 4 31B30.7B35.1 GB2 tok/s4K ctx
dense
DeepSeek V3.2671B411.5 GB2 tok/s4K ctx
moe
Nemotron Cascade 2 30B A3B30B22.9 GB7 tok/s4K ctx
moe
MiniMax M2.7
F0230B145.8 GB2 tok/s4K ctx
moe
Solar Open 100B
F0100B74.4 GB2 tok/s4K ctx
dense
Baichuan M2 32B Q4 K M
F032B25.0 GB2 tok/s4K ctx
dense
baichuan inc Baichuan M2 32B
F032B25.0 GB2 tok/s4K ctx
dense
Codestral 22B v0.1
F022B17.7 GB3 tok/s4K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
F024B19.2 GB2 tok/s4K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
F024B19.2 GB2 tok/s4K ctx
dense
internlm JanusCoder 14B
F014B11.9 GB12 tok/s4K ctx
dense
internlm2 5 20b chat
F020B16.2 GB4 tok/s4K ctx
dense
NousResearch Hermes 4 14B
F014B11.9 GB12 tok/s4K ctx
dense
starcoder2 15b instruct v0.1
F015B12.6 GB9 tok/s4K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
F048B36.6 GB2 tok/s4K ctx
dense
internlm2 limarp chat 20b
F020B16.2 GB4 tok/s4K ctx
dense
Phi 4 reasoning vision 15B
F015B12.6 GB9 tok/s4K ctx
dense
Codestral 22B v0.1 IMat
F022B17.7 GB3 tok/s4K ctx
dense
EXAONE 4.0 32B
F032B25.0 GB2 tok/s4K ctx
dense
K EXAONE 236B A23B
F0236B173.3 GB2 tok/s4K ctx
dense
Codestral 22B v0.1
F022B17.7 GB3 tok/s4K ctx
dense
Qwen3.5 35B A3B
F035B27.2 GB2 tok/s4K ctx
dense
starcoder2 15b instruct v0.1
F015B12.6 GB9 tok/s4K ctx
dense
gemma 3 12b it
F012B10.4 GB18 tok/s4K ctx
dense
gemma 3 27b it
F027B21.3 GB2 tok/s4K ctx
dense
Llama 3.3 70B Instruct
F070B52.6 GB2 tok/s4K ctx
dense
Mistral Small 24B Instruct 2501
F024B19.2 GB2 tok/s4K ctx
dense
mistral small 3.1 24b instruct 2503 hf
F024B19.2 GB2 tok/s4K ctx
dense
aya expanse 32b heretic MPOA i1
F032B25.0 GB2 tok/s4K ctx
dense
Baichuan M3 235B
F0235B172.6 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B172.6 GB2 tok/s4K ctx
dense
BaichuanMed OCR 72B i1
F072B54.1 GB2 tok/s4K ctx
dense
Codestral 21B Pruned i1
F021B17.0 GB3 tok/s4K ctx
dense
Codestral 22B v0.1 i1
F022B17.7 GB3 tok/s4K ctx
dense
Codestral RAG 19B Pruned i1
F019B15.5 GB5 tok/s4K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
F024B19.2 GB2 tok/s4K ctx
dense
HelpingAI 15B i1
F015B12.6 GB9 tok/s4K ctx
dense
internlm2 math plus 20b i1
F020B16.2 GB4 tok/s4K ctx
dense
Solar Open 100B i1
F0100B74.4 GB2 tok/s4K ctx
dense
Solar Open 69B REAP i1
F069B51.9 GB2 tok/s4K ctx
dense
starcoder2 15b i1
F015B12.6 GB9 tok/s4K ctx
dense
Hermes 4.3 36B
F036B27.9 GB2 tok/s4K ctx
dense
stabilityai japanese stablelm instruct beta 70b
F070B52.6 GB2 tok/s4K ctx
dense
Codestral 22B v0.1
F022B17.7 GB3 tok/s4K ctx
dense
StarCoder2 15B
F015B12.6 GB9 tok/s4K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
F014B11.9 GB12 tok/s4K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
F014B11.9 GB12 tok/s4K ctx
dense
DeepSeek R1 Distill Qwen 14B14B11.9 GB12 tok/s4K ctx
dense
gemma 3 27b it27B21.3 GB2 tok/s4K ctx
dense
Mistral Small 3.2 24B Instruct 250624B19.2 GB2 tok/s4K ctx
dense
Qwen3.5 122B A10B122B75.8 GB2 tok/s4K ctx
dense
Qwen3.5 27B27B21.3 GB2 tok/s4K ctx
dense
Qwen3.5 35B A3B35B27.2 GB2 tok/s4K ctx
dense
Qwen3.5 397B A17B397B290.4 GB2 tok/s4K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
F024B19.2 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs