AI Compatibility Tier List
NVIDIA DGX Spark 128GB
374 AI models ranked by how well they run on your 128 GB unified memory hardware. Best match: Qwen 3.5 122B A10B (score: 88).
374
Models tested
333
Can run
128 GB
unified memory
7
Best tok/s
Tier Distribution
S 10
A 56
B 70
C 197
F 41
S: 10A: 56B: 70C: 197F: 41
S
S Tier -- Perfect Fit (10 models)
These models run flawlessly with headroom to spare.
Qwen 3.5 122B A10B122B90.8 GB7 tok/s131K ctx
moe
Qwen3-Coder-Next80B64.2 GB11 tok/s256K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B34.0 GB25 tok/s256K ctx
moe
Qwen 3.6 35B A3B35B39.4 GB21 tok/s262K ctx
moe
Qwen3-VL 30B A3B Instruct30B33.7 GB26 tok/s256K ctx
moe
Command A 111B111B85.6 GB3 tok/s111K ctx
dense
Mistral Small 4 119B119B91.9 GB7 tok/s66K ctx
moe
Qwen 3.5 35B A3B35B36.8 GB23 tok/s131K ctx
moe
Devstral 2 123B Instruct123B94.4 GB2 tok/s59K ctx
dense
Qwen 2.5 VL 72B72B62.8 GB4 tok/s33K ctx
dense
A
A Tier -- Great Fit (56 models)
Excellent performance with comfortable memory margins.
Qwen 3 30B A3B30.5B34.0 GB25 tok/s131K ctx
moe
Qwen 3.5 9B9B21.6 GB32 tok/s131K ctx
dense
Qwen 3.5 27B27B33.6 GB11 tok/s131K ctx
dense
Qwen 3.5 4B4B18.6 GB56 tok/s131K ctx
dense
GPT-OSS 120B117B90.2 GB3 tok/s77K ctx
dense
Magistral Small 250724B31.0 GB12 tok/s131K ctx
dense
Qwen 3 8B8B21.0 GB36 tok/s131K ctx
dense
Qwen 3 32B32B37.4 GB9 tok/s131K ctx
dense
Qwen 3.6 27B27B31.4 GB8 tok/s262K ctx
dense
Devstral Small 2 24B Instruct24B31.0 GB12 tok/s256K ctx
dense
GPT-OSS 20B21B29.2 GB31 tok/s128K ctx
moe
Qwen 3 14B14B24.9 GB21 tok/s131K ctx
dense
Pixtral Large 124B124B95.0 GB2 tok/s57K ctx
dense
Nemotron Cascade 2 30B A3B30B35.2 GB25 tok/s262K ctx
moe
Phi-4-reasoning-plus 14B14.7B26.0 GB20 tok/s33K ctx
dense
Nemotron 3 Nano 30B30B34.7 GB10 tok/s131K ctx
dense
Leanstral 119B A6B119B95.3 GB7 tok/s41K ctx
moe
Devstral Small 1.124B31.0 GB12 tok/s131K ctx
dense
Phi-4 Mini Reasoning 4B3.8B17.7 GB53 tok/s131K ctx
dense
Gemma 4 31B30.7B47.3 GB7 tok/s83K ctx
dense
Gemma 4 26B A4B25.2B33.0 GB27 tok/s256K ctx
moe
Llama 3.3 70B70B61.5 GB4 tok/s128K ctx
dense
Nemotron Nano 8B8B20.8 GB36 tok/s131K ctx
dense
Ministral 3 14B14B24.9 GB21 tok/s262K ctx
multimodal
EXAONE 4.0 32B32B37.4 GB9 tok/s131K ctx
dense
Llama 4 Scout 17B 16E109B83.4 GB6 tok/s155K ctx
moe
Qwen 2.5 72B72B62.8 GB4 tok/s131K ctx
dense
Llama 3.1 70B70B61.5 GB4 tok/s128K ctx
dense
Qwen 2.5 32B32B37.4 GB9 tok/s131K ctx
dense
Codestral 2 25.0822B29.8 GB12 tok/s256K ctx
dense
Mistral Small 3.2 24B24B31.0 GB12 tok/s131K ctx
vision
Qwen 3 4B4B18.6 GB56 tok/s33K ctx
dense
InternVL2 8B8B20.8 GB36 tok/s8K ctx
dense
CogVLM2 19B19B28.0 GB15 tok/s8K ctx
dense
Gemma 3 27B27B41.7 GB8 tok/s112K ctx
dense
LFM2 24B24B31.0 GB12 tok/s131K ctx
dense
Jina Embeddings v30.57B17.1 GB8 tok/s8K ctx
dense
OLMo 2 32B32B37.4 GB9 tok/s4K ctx
dense
Kimi Linear 48B A3B48B44.2 GB6 tok/s1049K ctx
linear
Granite 4.1 30B30B36.2 GB10 tok/s131K ctx
dense
Phi-4 14B14B25.5 GB21 tok/s16K ctx
dense
Ministral 3 8B8B21.0 GB36 tok/s262K ctx
multimodal
MiniCPM-V 2.6 8B8B20.8 GB36 tok/s2K ctx
dense
Mistral Small 24B24B31.0 GB12 tok/s33K ctx
dense
BGE M30.57B16.2 GB8 tok/s8K ctx
dense
DeepSeek Coder V2 16B16B27.0 GB40 tok/s131K ctx
moe
SQLCoder 7B7B20.2 GB41 tok/s8K ctx
dense
Qwen 2.5 14B14B25.4 GB21 tok/s131K ctx
dense
Mistral Small 3.1 24B24B31.0 GB12 tok/s131K ctx
dense
Qwen 2.5 VL 7B7B19.1 GB42 tok/s33K ctx
dense
Nemotron Nano 9B v29B21.9 GB32 tok/s131K ctx
dense
Gemma 3 12B12B26.2 GB18 tok/s131K ctx
dense
Granite Code 20B20B29.3 GB15 tok/s8K ctx
dense
Magistral 7B7B20.2 GB41 tok/s8K ctx
dense
DeepSeek R1 Distill 70B70B61.5 GB4 tok/s131K ctx
dense
CodeGeeX 4 9B9B20.1 GB33 tok/s131K ctx
dense
B
B Tier -- Good Fit (70 models)
Solid performance, may need quantization or reduced context.
mxbai Embed Large0.34B16.1 GB5 tok/s1K ctx
dense
Qwen 2.5 Coder 32B32B37.4 GB9 tok/s131K ctx
dense
Gemma 4 E4B8B20.1 GB27 tok/s128K ctx
dense
StarCoder 15B15B39.4 GB16 tok/s8K ctx
dense
Snowflake Arctic Embed L0.34B16.1 GB5 tok/s1K ctx
dense
OLMo 2 13B13B24.3 GB22 tok/s33K ctx
dense
Qwen 2.5 7B7B19.1 GB42 tok/s131K ctx
dense
CodeLlama 13B Instruct13B34.1 GB21 tok/s16K ctx
dense
Granite Code 34B34B38.4 GB9 tok/s8K ctx
dense
Qwen 2.5 Coder 3B3B18.0 GB42 tok/s131K ctx
dense
DevStral 7B7B20.2 GB41 tok/s8K ctx
dense
Codestral Mamba 7B7B18.7 GB44 tok/s262K ctx
state-space
Granite Code 8B8B20.8 GB36 tok/s8K ctx
dense
Command R 35B35B37.7 GB8 tok/s131K ctx
dense
CodeLlama 7B Instruct7B26.0 GB38 tok/s16K ctx
dense
StarCoder 7B7B25.6 GB38 tok/s8K ctx
dense
Granite 4.1 8B8B21.3 GB36 tok/s131K ctx
dense
DeepSeek R1 Distill 32B32B37.4 GB9 tok/s33K ctx
dense
BGE Large EN v1.50.34B16.1 GB5 tok/s1K ctx
dense
LLaVA 1.6 13B13B34.1 GB21 tok/s4K ctx
dense
DeepSeek R1 Distill 14B14B25.4 GB21 tok/s33K ctx
dense
MPT-30B-Instruct30B59.0 GB8 tok/s8K ctx
dense
Nemotron 70B70B61.5 GB4 tok/s131K ctx
dense
Pixtral 12B12B23.7 GB24 tok/s131K ctx
dense
Gemma 4 E2B5.1B17.6 GB43 tok/s128K ctx
dense
Nomic Embed Text v1.50.14B14.8 GB2 tok/s8K ctx
dense
Nous Hermes 1.09B31.7 GB30 tok/s16K ctx
dense
Gemma 3 4B4B18.5 GB53 tok/s128K ctx
dense
Ministral 3 3B3B16.5 GB42 tok/s262K ctx
multimodal
InternLM 7B7B26.0 GB38 tok/s8K ctx
dense
InternLM Chat 7B7B26.0 GB38 tok/s8K ctx
dense
WizardLM 13B13B34.1 GB21 tok/s8K ctx
dense
Vicuna 13B13B34.1 GB21 tok/s4K ctx
dense
Phi 4 Mini 4B4B17.9 GB56 tok/s128K ctx
dense
Nous Dolphin 13B13B35.5 GB18 tok/s16K ctx
dense
Solar 7B7B21.2 GB41 tok/s8K ctx
dense
Llama 3.1 8B8B20.8 GB36 tok/s128K ctx
dense
WizardMath 7B7B20.2 GB41 tok/s4K ctx
dense
OLMo 2 7B7B20.2 GB41 tok/s4K ctx
dense
GLM-4 9B9B20.1 GB33 tok/s128K ctx
dense
Qwen 3.5 2B2B16.9 GB28 tok/s131K ctx
dense
Command R+ 104B104B80.8 GB3 tok/s131K ctx
dense
Qwen 2.5 Coder 7B7B19.1 GB42 tok/s131K ctx
dense
Falcon 40B Instruct40B44.6 GB6 tok/s8K ctx
dense
LLaVA 1.5 7B7B26.0 GB38 tok/s4K ctx
dense
Gemma 2 27B27B41.7 GB8 tok/s8K ctx
dense
Qwen 2.5 3B3B18.0 GB42 tok/s131K ctx
dense
Phi 3 Mini 3.8B3.8B22.1 GB53 tok/s128K ctx
dense
Qwen 3 1.7B1.7B16.7 GB24 tok/s33K ctx
dense
Phi 3.5 Mini 4B4B22.3 GB56 tok/s128K ctx
dense
MPT-7B-Instruct7B26.0 GB38 tok/s8K ctx
dense
Baichuan 7B7B26.0 GB38 tok/s8K ctx
dense
All MiniLM L6 v20.02B14.3 GB2 tok/s0K ctx
dense
DeepSeek R1 Distill 7B7B19.1 GB42 tok/s33K ctx
dense
Baichuan 13B13B35.5 GB18 tok/s8K ctx
dense
Granite 4.1 3B3B17.0 GB42 tok/s131K ctx
dense
Mixtral 8x7B47B44.6 GB12 tok/s33K ctx
moe
DeepSeek R1 Distill 8B8B20.8 GB36 tok/s33K ctx
dense
Samantha 7B7B20.2 GB41 tok/s4K ctx
dense
Cerebras-GPT 13B13B33.1 GB18 tok/s131K ctx
dense
Falcon 7B Instruct7B18.3 GB42 tok/s8K ctx
dense
Granite Code 3B3B18.3 GB42 tok/s8K ctx
dense
Qwen 2.5 Coder 1.5B1.5B15.3 GB21 tok/s33K ctx
dense
Qwen 2.5 Math 72B72B62.8 GB4 tok/s4K ctx
dense
Mixtral 8x22B141B103.4 GB4 tok/s41K ctx
moe
Gemma 2 9B9B24.6 GB24 tok/s8K ctx
dense
Llama 3.2 11B Vision11B22.6 GB26 tok/s16K ctx
vision
Qwen 2.5 Coder 14B14B25.4 GB21 tok/s131K ctx
dense
Mistral 7B Instruct v0.37B20.2 GB41 tok/s8K ctx
dense
Llama 3.2 3B3B17.5 GB42 tok/s128K ctx
dense
C
C Tier -- Tight Fit (197 models)
Runs with offloading or significant compromises.
Yi Coder 9B9B20.9 GB32 tok/s131K ctx
dense
Mistral Nemo 12B12B23.7 GB24 tok/s128K ctx
dense
Yi 1.5 34B34B38.4 GB9 tok/s4K ctx
dense
DeepSeek LLM 67B67B60.6 GB4 tok/s4K ctx
dense
Phi 3 Medium 14B14B25.5 GB21 tok/s128K ctx
dense
Ministral 8B8B21.0 GB36 tok/s131K ctx
dense
InternLM 20B20B46.7 GB12 tok/s8K ctx
dense
Codestral 22B22B29.8 GB13 tok/s33K ctx
dense
DeepSeek R1 1.5B1.5B15.3 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B15.0 GB15 tok/s4K ctx
dense
SmolLM3 3B3B17.7 GB42 tok/s128K ctx
dense
Qwen 2.5 1.5B1.5B15.3 GB21 tok/s131K ctx
dense
Granite 3.1 8B8B20.8 GB42 tok/s128K ctx
state-space
Qwen 2.5 Math 7B7B19.1 GB42 tok/s4K ctx
dense
Yi 1.5 9B9B20.9 GB32 tok/s4K ctx
dense
Aya Expanse 32B32B35.9 GB9 tok/s8K ctx
dense
Gemma 3 1B1B15.0 GB14 tok/s33K ctx
dense
Qwen3.5 122B A10B122B88.0 GB3 tok/s39K ctx
dense
Gemma 2 2B2B16.8 GB28 tok/s8K ctx
dense
Solar Open 100B
C46100B86.7 GB3 tok/s46K ctx
dense
Solar Open 100B i1
C46100B86.7 GB3 tok/s46K ctx
dense
Qwen 2.5 Coder 0.5B0.5B14.4 GB7 tok/s131K ctx
dense
OpenChat 7B7B20.2 GB41 tok/s8K ctx
dense
Llama 3.3 70B Instruct
C4570B64.9 GB4 tok/s102K ctx
dense
Vicuna 7B7B26.0 GB38 tok/s4K ctx
dense
Nemotron Mini 4B4B18.3 GB56 tok/s4K ctx
dense
Solar Open 69B REAP i1
C4569B64.1 GB4 tok/s104K ctx
dense
Aya Expanse 8B8B20.8 GB36 tok/s8K ctx
dense
BaichuanMed OCR 72B i1
C4472B66.3 GB4 tok/s97K ctx
dense
stabilityai japanese stablelm instruct beta 70b
C4470B64.9 GB4 tok/s102K ctx
dense
OpenHermes 2.5 7B7B20.2 GB41 tok/s8K ctx
dense
Qwen3.5 4B4B16.9 GB56 tok/s3154K ctx
dense
gemma 3 4b it
C444B16.9 GB56 tok/s3154K ctx
dense
Zephyr 7B Beta7B20.2 GB41 tok/s33K ctx
dense
gemma 3 4b it
C444B16.9 GB56 tok/s3154K ctx
dense
Dolphin 2.9 8B8B20.8 GB36 tok/s33K ctx
dense
Starling LM 7B7B20.2 GB41 tok/s8K ctx
dense
Qwen 3 0.6B0.6B15.2 GB8 tok/s33K ctx
dense
Yi 34B Chat34B38.4 GB9 tok/s200K ctx
dense
Llama 3.2 1B1B15.1 GB14 tok/s128K ctx
dense
Yi 1.5 6B6B18.6 GB49 tok/s4K ctx
dense
Yi 1.5 6B Chat
C436B18.3 GB45 tok/s2075K ctx
dense
StableLM 2 12B12B34.8 GB18 tok/s4K ctx
dense
stablelm 2 zephyr 1 6b
C436B18.3 GB45 tok/s2075K ctx
dense
HelpingAI2.5 5B i1
C435B17.6 GB54 tok/s2507K ctx
dense
Llama 2 7B Chat
C437B19.0 GB38 tok/s1767K ctx
dense
falcon mamba 7b instruct Q4 K M
C437B19.0 GB44 tok/s1767K ctx
dense
Mistral 7B Instruct v0.2
C437B19.0 GB38 tok/s1767K ctx
dense
Mistral 7B Instruct v0.3
C437B19.0 GB38 tok/s1767K ctx
dense
Yi 1.5 6B Chat
C436B18.3 GB45 tok/s2075K ctx
dense
stablelm 2 1 6b chat imatrix
C436B18.3 GB45 tok/s2075K ctx
dense
Llama 3.2 3B Instruct
C433B16.5 GB42 tok/s4218K ctx
dense
StarCoder2 15B15B26.0 GB17 tok/s16K ctx
dense
Mamba Codestral 7B v0.1
C437B19.0 GB44 tok/s1767K ctx
dense
Meta Llama 3.1 8B Instruct
C438B19.8 GB34 tok/s1535K ctx
dense
llava llama 3 8b v1 1
C438B19.8 GB34 tok/s1535K ctx
dense
Qwen2.5 3B Instruct3B16.1 GB42 tok/s4233K ctx
dense
Mistral 7B Instruct v0.3
C437B19.0 GB38 tok/s1767K ctx
dense
Hermes 2 Pro Mistral 7B
C437B19.0 GB38 tok/s1767K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B19.8 GB34 tok/s1535K ctx
dense
zephyr 7B beta
C437B19.0 GB38 tok/s1767K ctx
dense
Meta Llama 3 8B Instruct
C438B19.8 GB34 tok/s1535K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C437B19.0 GB38 tok/s1767K ctx
dense
DeepSeek R1 Distill Llama 8B8B19.8 GB34 tok/s1535K ctx
dense
Dolphin3.0 Llama3.1 8B
C438B19.8 GB34 tok/s1535K ctx
dense
Llama 3 8B Instruct 32k v0.1
C438B19.8 GB34 tok/s1535K ctx
dense
HelpingAI2 6B
C436B18.3 GB45 tok/s2075K ctx
dense
Falcon H1R 7B
C437B19.0 GB38 tok/s1767K ctx
dense
Meta Llama 3.1 8B Instruct
C438B19.8 GB34 tok/s1535K ctx
dense
vntl llama3 8b v2
C438B19.8 GB34 tok/s1535K ctx
dense
Qwen3.5 9B9B20.5 GB30 tok/s1356K ctx
dense
HelpingAI2 6B i1
C436B18.3 GB45 tok/s2075K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C438B19.8 GB34 tok/s1535K ctx
dense
zephyr 7B alpha
C437B19.0 GB38 tok/s1767K ctx
dense
speechless zephyr code functionary 7b
C437B19.0 GB38 tok/s1767K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C439B20.5 GB30 tok/s1356K ctx
dense
Llama 3.2 3B Instruct
C433B16.1 GB42 tok/s4233K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C438B19.8 GB34 tok/s1535K ctx
dense
StarCoder2 7B
C437B19.0 GB38 tok/s1767K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C4348B48.9 GB6 tok/s186K ctx
dense
Ministral 3 3B Instruct 2512
C433B16.1 GB42 tok/s4233K ctx
dense
japanese stablelm instruct gamma 7B
C437B19.0 GB38 tok/s1767K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C438B19.8 GB34 tok/s1535K ctx
dense
Falcon H1 7B Instruct
C437B19.0 GB38 tok/s1767K ctx
dense
Hermes 3 Llama 3.1 8B
C438B19.8 GB34 tok/s1535K ctx
dense
Falcon H1R 7B7B19.0 GB38 tok/s1767K ctx
dense
Neural Chat 7B7B20.2 GB41 tok/s8K ctx
dense
Qwen3.5 9B
C439B20.5 GB30 tok/s1356K ctx
dense
Hermes 2 Pro Llama 3 8B
C428B19.8 GB34 tok/s1535K ctx
dense
Dolphin3.0 Llama3.1 8B
C428B19.8 GB34 tok/s1535K ctx
dense
DeepSeek LLM 7B7B25.6 GB38 tok/s4K ctx
dense
granite 8b code instruct 4k
C428B19.8 GB34 tok/s1535K ctx
dense
stabilityai japanese stablelm base gamma 7b
C427B19.0 GB38 tok/s1767K ctx
dense
dolphin 2.9.4 llama3.1 8b
C428B19.8 GB34 tok/s1535K ctx
dense
HELVETE 3B
C423B16.1 GB42 tok/s4233K ctx
dense
baichuan2 7b chat
C427B19.0 GB38 tok/s1767K ctx
dense
DiscoPOP zephyr 7b gemma
C427B19.0 GB38 tok/s1767K ctx
dense
EXAONE 3.5 7.8B Instruct
C427.8B19.6 GB34 tok/s1577K ctx
dense
Hermes 3 Llama 3.2 3B
C423B16.1 GB42 tok/s4233K ctx
dense
starcoder2 7b
C427B19.0 GB38 tok/s1767K ctx
dense
stablelm zephyr 3b
C423B16.1 GB42 tok/s4233K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C427B19.0 GB38 tok/s1767K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C427B19.0 GB38 tok/s1767K ctx
dense
openchat 3.6 8b 20240522 IMat
C428B19.8 GB34 tok/s1535K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C427B19.0 GB38 tok/s1767K ctx
dense
internlm2 math plus 7b IMat
C427B19.0 GB38 tok/s1767K ctx
dense
Yi Coder 9B Chat
C429B20.5 GB30 tok/s1356K ctx
dense
StarCoder2 3B
C423B16.1 GB42 tok/s4233K ctx
dense
EXAONE 3.5 7.8B Instruct
C427.8B19.6 GB34 tok/s1577K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C427B19.0 GB38 tok/s1767K ctx
dense
glm 4 9b chat 1m
C429B20.5 GB30 tok/s1356K ctx
dense
aya expanse 8b
C428B19.8 GB34 tok/s1535K ctx
dense
internlm2 5 7b chat i1
C427B19.0 GB38 tok/s1767K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C427.8B19.6 GB34 tok/s1577K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C427B19.0 GB38 tok/s1767K ctx
dense
MD Judge v0 2 internlm2 7b i1
C427B19.0 GB38 tok/s1767K ctx
dense
dolphin v2 8b abliterated i1
C428B19.8 GB34 tok/s1535K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C427B19.0 GB38 tok/s1767K ctx
dense
exaone 3.0 7.8b it
C427.8B19.6 GB34 tok/s1577K ctx
dense
jointpreferences mistral 7b sft helpful
C427B19.0 GB38 tok/s1767K ctx
dense
zephyr 7b dpo full i1
C427B19.0 GB38 tok/s1767K ctx
dense
blossom v3 baichuan2 7b i1
C427B19.0 GB38 tok/s1767K ctx
dense
AI21 Jamba Reasoning 3B
C423B16.1 GB42 tok/s4233K ctx
dense
blossom v1 baichuan 7b i1
C427B19.0 GB38 tok/s1767K ctx
dense
StarCoder2 7B7B18.7 GB42 tok/s16K ctx
dense
aya expanse 8b orthogonal heretic i1
C428B19.8 GB34 tok/s1535K ctx
dense
Qwen3.5 35B A3B35B39.4 GB8 tok/s287K ctx
dense
Yi 1.5 9B Chat
C429B20.5 GB30 tok/s1356K ctx
dense
stablelm 3b 4e1t
C423B16.1 GB42 tok/s4233K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C423B16.1 GB42 tok/s4233K ctx
dense
aya expanse 8b orthogonal heretic
C428B19.8 GB34 tok/s1535K ctx
dense
ai21labs AI21 Jamba2 3B
C423B16.1 GB42 tok/s4233K ctx
dense
internlm2 5 1 8b chat i1
C428B19.8 GB34 tok/s1535K ctx
dense
internlm3 8b instruct abliterated i1
C428B19.8 GB34 tok/s1535K ctx
dense
HelpingAI 3B hindi i1
C423B16.1 GB42 tok/s4233K ctx
dense
AI21 Jamba2 3B
C423B16.1 GB42 tok/s4233K ctx
dense
Qwen3.5 35B A3B
C4235B39.4 GB8 tok/s287K ctx
dense
HelpingAI 3B hindi
C423B16.1 GB42 tok/s4233K ctx
dense
AI21 Jamba2 3B i1
C423B16.1 GB42 tok/s4233K ctx
dense
Yi 9B Coder i1
C429B20.5 GB30 tok/s1356K ctx
dense
gemma 3 12b it
C4212B22.7 GB22 tok/s996K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4210.7B21.7 GB25 tok/s1127K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4210.7B21.7 GB25 tok/s1127K ctx
dense
HelpingAI2 9B
C429B20.5 GB30 tok/s1356K ctx
dense
HelpingAI2 9B i1
C429B20.5 GB30 tok/s1356K ctx
dense
Qwen3.5 27B27B33.6 GB10 tok/s396K ctx
dense
HelpingAI 9B 200k i1
C429B20.5 GB30 tok/s1356K ctx
dense
HelpingAI 9B i1
C429B20.5 GB30 tok/s1356K ctx
dense
DeepSeek R1 Distill Qwen 14B14B24.1 GB19 tok/s842K ctx
dense
Hermes 4.3 36B
C4236B40.1 GB8 tok/s276K ctx
dense
gemma 3 27b it27B33.6 GB10 tok/s396K ctx
dense
solar finalised finetuned Model 10.7B i1
C4210.7B21.7 GB25 tok/s1127K ctx
dense
HelpingAI2.5 10B i1
C4210B21.2 GB27 tok/s1212K ctx
dense
SOLAR 10.7B v1.0
C4210.7B21.7 GB25 tok/s1127K ctx
dense
EXAONE 3.5 2.4B Instruct
C422.4B15.7 GB34 tok/s5312K ctx
dense
Helply 10.2b chat i1
C4210.2B21.4 GB26 tok/s1186K ctx
dense
gemma 2 2b it
C422B15.8 GB28 tok/s6363K ctx
dense
Mistral Small 3.2 24B Instruct 250624B31.4 GB11 tok/s456K ctx
dense
gemma 3 27b it
C4127B33.6 GB10 tok/s396K ctx
dense
NousResearch Hermes 4 14B
C4114B24.1 GB19 tok/s842K ctx
dense
gemma 2b2B15.4 GB28 tok/s6391K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4124B31.4 GB11 tok/s456K ctx
dense
Phi 4 reasoning vision 15B
C4115B24.9 GB18 tok/s780K ctx
dense
Mistral Small 24B Instruct 2501
C4124B31.4 GB11 tok/s456K ctx
dense
StarCoder2 15B
C4115B24.9 GB18 tok/s780K ctx
dense
EXAONE 4.0 32B
C4132B37.2 GB8 tok/s321K ctx
dense
aya expanse 32b heretic MPOA i1
C4132B37.2 GB8 tok/s321K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4124B31.4 GB11 tok/s456K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4124B31.4 GB11 tok/s456K ctx
dense
Gemmasutra Mini 2B v1
C412B15.4 GB28 tok/s6391K ctx
dense
Baichuan M2 32B Q4 K M
C4132B37.2 GB8 tok/s321K ctx
dense
Codestral 22B v0.1
C4122B30.0 GB12 tok/s505K ctx
dense
Codestral 22B v0.1
C4122B30.0 GB12 tok/s505K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4124B31.4 GB11 tok/s456K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4114B24.1 GB19 tok/s842K ctx
dense
baichuan inc Baichuan M2 32B
C4132B37.2 GB8 tok/s321K ctx
dense
internlm JanusCoder 14B
C4114B24.1 GB19 tok/s842K ctx
dense
starcoder2 15b instruct v0.1
C4115B24.9 GB18 tok/s780K ctx
dense
Qwen 3.5 0.6B0.6B15.2 GB8 tok/s131K ctx
dense
starcoder2 15b i1
C4115B24.9 GB18 tok/s780K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4114B24.1 GB19 tok/s842K ctx
dense
gemma 2 2b it
C412B15.4 GB28 tok/s6391K ctx
dense
starcoder2 15b instruct v0.1
C4115B24.9 GB18 tok/s780K ctx
dense
Codestral 22B v0.1
C4122B30.0 GB12 tok/s505K ctx
dense
StarCoder2 3B3B16.2 GB42 tok/s16K ctx
dense
HelpingAI 15B i1
C4115B24.9 GB18 tok/s780K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4124B31.4 GB11 tok/s456K ctx
dense
Codestral 22B v0.1 IMat
C4122B30.0 GB12 tok/s505K ctx
dense
internlm2 5 20b chat
C4120B28.5 GB13 tok/s564K ctx
dense
internlm2 limarp chat 20b
C4120B28.5 GB13 tok/s564K ctx
dense
Codestral 21B Pruned i1
C4121B29.2 GB13 tok/s533K ctx
dense
internlm2 math plus 20b i1
C4120B28.5 GB13 tok/s564K ctx
dense
Codestral RAG 19B Pruned i1
C4119B27.8 GB14 tok/s598K ctx
dense
Codestral 22B v0.1 i1
C4122B30.0 GB12 tok/s505K ctx
dense
Qwen2.5 1.5B Instruct1.5B15.0 GB21 tok/s8550K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B15.0 GB21 tok/s8550K ctx
dense
Yi Coder 1.5B Chat
C401.5B15.0 GB21 tok/s8550K ctx
dense
F
F Tier -- Can't Run (41 models)
Too heavy for this hardware.
stablelm 2 zephyr 1.6b
D401.6B15.1 GB22 tok/s8010K ctx
dense
logos16v2 stablelm2 1.6b i1
D401.6B15.1 GB22 tok/s8010K ctx
dense
Yi Coder 1.5B
D401.5B15.0 GB21 tok/s8550K ctx
dense
Falcon H1 1.5B Instruct1.5B15.0 GB21 tok/s8550K ctx
dense
TinyLlama 1.1B Chat v1.0
D391.1B14.8 GB15 tok/s11689K ctx
dense
Qwen 2.5 0.5B0.5B14.4 GB7 tok/s131K ctx
dense
EXAONE 4.0 1.2B
D391.2B14.8 GB17 tok/s10708K ctx
dense
Llama 3.2 1B Instruct Q8 0
D391B14.9 GB14 tok/s12837K ctx
dense
TinyLlama 1.1B Chat v0.3
D391.1B14.8 GB15 tok/s11689K ctx
dense
TinyLlama 1.1B Chat v0.61.1B14.8 GB15 tok/s11689K ctx
dense
Llama 3.2 1B Instruct
D391B14.7 GB14 tok/s12866K ctx
dense
gemma 3 1b it
D391B14.7 GB14 tok/s12866K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D391.1B14.8 GB15 tok/s11689K ctx
dense
Falcon3 1B Instruct abliterated
D381B14.7 GB14 tok/s12866K ctx
dense
SmolVLM 500M Instruct
D370.5B14.5 GB7 tok/s15109K ctx
dense
embeddinggemma 300M
D350.3B14.3 GB4 tok/s15136K ctx
dense
granite embedding 107m multilingual
D330.11B14.1 GB2 tok/s15165K ctx
dense
Falcon H1 Tiny 90M Instruct
D330.09B14.1 GB2 tok/s15166K ctx
dense
Qwen 3 235B A22B235B160.2 GB2 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B309.6 GB2 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B259.0 GB2 tok/s4K ctx
moe
Mistral Large 3675B431.1 GB2 tok/s4K ctx
moe
GLM-5744B486.8 GB2 tok/s4K ctx
moe
GLM-5.1754B492.9 GB2 tok/s4K ctx
moe
Kimi K2.51000B631.4 GB2 tok/s4K ctx
moe
Kimi K2.61000B631.4 GB2 tok/s4K ctx
moe
DeepSeek V3 671B671B482.8 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B482.8 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B877.8 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B173.3 GB2 tok/s4K ctx
moe
DeepSeek R1 671B671B482.8 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B216.5 GB2 tok/s4K ctx
moe
DeepSeek V2.5 236B236B216.5 GB2 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B260.9 GB2 tok/s4K ctx
moe
Llama 3.1 405B405B268.7 GB2 tok/s4K ctx
dense
DeepSeek V3.2671B423.7 GB2 tok/s4K ctx
moe
MiniMax M2.7
F0230B158.0 GB2 tok/s4K ctx
moe
K EXAONE 236B A23B
F0236B185.6 GB2 tok/s4K ctx
dense
Baichuan M3 235B
F0235B184.8 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B184.8 GB2 tok/s4K ctx
dense
Qwen3.5 397B A17B397B302.6 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs