AI Compatibility Tier List
RX 7600 XT 16GB
374 AI models ranked by how well they run on your 16 GB VRAM hardware. Best match: Qwen 3.5 9B (score: 94).
374
Models tested
246
Can run
16 GB
VRAM
33
Best tok/s
Tier Distribution
S 7
A 45
B 31
C 163
F 128
S: 7A: 45B: 31C: 163F: 128
S
S Tier -- Perfect Fit (7 models)
These models run flawlessly with headroom to spare.
Qwen 3.5 9B9B10.2 GB33 tok/s58K ctx
dense
Qwen 3 8B8B9.6 GB37 tok/s63K ctx
dense
Qwen 3.5 4B4B7.1 GB56 tok/s81K ctx
dense
Qwen 3 14B14B13.5 GB21 tok/s33K ctx
dense
Phi-4-reasoning-plus 14B14.7B14.5 GB20 tok/s24K ctx
dense
Nemotron Nano 8B8B9.3 GB37 tok/s71K ctx
dense
Phi-4 Mini Reasoning 4B3.8B6.3 GB53 tok/s122K ctx
dense
A
A Tier -- Great Fit (45 models)
Excellent performance with comfortable memory margins.
Ministral 3 14B14B13.5 GB21 tok/s33K ctx
multimodal
InternVL2 8B8B9.3 GB37 tok/s8K ctx
dense
Ministral 3 8B8B9.6 GB37 tok/s63K ctx
multimodal
Nemotron Nano 9B v29B10.4 GB33 tok/s52K ctx
dense
MiniCPM-V 2.6 8B8B9.3 GB37 tok/s2K ctx
dense
Qwen 3 4B4B7.1 GB56 tok/s33K ctx
dense
Phi-4 14B14B14.1 GB21 tok/s16K ctx
dense
SQLCoder 7B7B8.7 GB42 tok/s8K ctx
dense
CogVLM2 19B19B16.5 GB11 tok/s8K ctx
dense
Qwen 2.5 14B14B14.0 GB21 tok/s27K ctx
dense
DeepSeek Coder V2 16B16B15.6 GB41 tok/s18K ctx
moe
OLMo 2 13B13B12.9 GB23 tok/s33K ctx
dense
Qwen 2.5 VL 7B7B7.6 GB43 tok/s33K ctx
dense
Magistral 7B7B8.7 GB42 tok/s8K ctx
dense
Jina Embeddings v30.57B5.6 GB8 tok/s8K ctx
dense
Gemma 3 12B12B14.7 GB18 tok/s20K ctx
dense
CodeGeeX 4 9B9B8.6 GB33 tok/s131K ctx
dense
GPT-OSS 20B21B17.8 GB19 tok/s5K ctx
moe
Gemma 4 E4B8B8.7 GB28 tok/s108K ctx
dense
BGE M30.57B4.8 GB8 tok/s8K ctx
dense
Granite Code 8B8B9.3 GB37 tok/s8K ctx
dense
Granite 4.1 8B8B9.8 GB37 tok/s56K ctx
dense
Pixtral 12B12B12.3 GB25 tok/s41K ctx
dense
DevStral 7B7B8.7 GB42 tok/s8K ctx
dense
Qwen 2.5 7B7B7.6 GB43 tok/s131K ctx
dense
Codestral Mamba 7B7B7.3 GB45 tok/s262K ctx
state-space
StarCoder 7B7B14.1 GB39 tok/s8K ctx
dense
CodeLlama 7B Instruct7B14.6 GB39 tok/s16K ctx
dense
Qwen 2.5 Coder 3B3B6.5 GB42 tok/s85K ctx
dense
DeepSeek R1 Distill 14B14B14.0 GB21 tok/s27K ctx
dense
mxbai Embed Large0.34B4.7 GB5 tok/s1K ctx
dense
Solar 7B7B9.7 GB42 tok/s8K ctx
dense
Llama 3.1 8B8B9.3 GB37 tok/s71K ctx
dense
Snowflake Arctic Embed L0.34B4.7 GB5 tok/s1K ctx
dense
InternLM 7B7B14.6 GB39 tok/s8K ctx
dense
InternLM Chat 7B7B14.6 GB39 tok/s8K ctx
dense
WizardMath 7B7B8.7 GB42 tok/s4K ctx
dense
OLMo 2 7B7B8.7 GB42 tok/s4K ctx
dense
GLM-4 9B9B8.6 GB33 tok/s128K ctx
dense
Phi 3 Mini 3.8B3.8B10.7 GB53 tok/s31K ctx
dense
Gemma 3 4B4B7.0 GB55 tok/s85K ctx
dense
Gemma 4 E2B5.1B6.1 GB44 tok/s128K ctx
dense
Phi 3.5 Mini 4B4B10.8 GB56 tok/s30K ctx
dense
BGE Large EN v1.50.34B4.7 GB5 tok/s1K ctx
dense
Codestral 2 25.0822B18.4 GB7 tok/s4K ctx
dense
B
B Tier -- Good Fit (31 models)
Solid performance, may need quantization or reduced context.
Qwen 2.5 Coder 7B7B7.6 GB43 tok/s131K ctx
dense
Phi 4 Mini 4B4B6.4 GB56 tok/s121K ctx
dense
LLaVA 1.5 7B7B14.6 GB39 tok/s4K ctx
dense
Ministral 3 3B3B5.1 GB42 tok/s255K ctx
multimodal
DeepSeek R1 Distill 8B8B9.3 GB37 tok/s33K ctx
dense
Qwen 3.5 2B2B5.4 GB28 tok/s115K ctx
dense
Llama 3.2 11B Vision11B11.2 GB27 tok/s16K ctx
vision
Nomic Embed Text v1.50.14B3.3 GB2 tok/s8K ctx
dense
MPT-7B-Instruct7B14.6 GB39 tok/s8K ctx
dense
Baichuan 7B7B14.6 GB39 tok/s8K ctx
dense
Samantha 7B7B8.7 GB42 tok/s4K ctx
dense
Gemma 2 9B9B13.1 GB24 tok/s8K ctx
dense
Qwen 2.5 3B3B6.5 GB42 tok/s85K ctx
dense
DeepSeek R1 Distill 7B7B7.6 GB43 tok/s33K ctx
dense
Granite Code 20B20B17.9 GB9 tok/s7K ctx
dense
Mistral Nemo 12B12B12.3 GB25 tok/s41K ctx
dense
Falcon 7B Instruct7B6.9 GB43 tok/s8K ctx
dense
Granite Code 3B3B6.8 GB42 tok/s8K ctx
dense
Qwen 3 1.7B1.7B5.2 GB24 tok/s33K ctx
dense
Granite 4.1 3B3B5.6 GB42 tok/s131K ctx
dense
Yi Coder 9B9B9.5 GB33 tok/s87K ctx
dense
Qwen 2.5 Coder 14B14B14.0 GB21 tok/s27K ctx
dense
Mistral 7B Instruct v0.37B8.7 GB42 tok/s8K ctx
dense
Ministral 8B8B9.6 GB37 tok/s63K ctx
dense
All MiniLM L6 v20.02B2.8 GB2 tok/s0K ctx
dense
Qwen 2.5 Coder 1.5B1.5B3.8 GB21 tok/s33K ctx
dense
Phi 3 Medium 14B14B14.1 GB21 tok/s26K ctx
dense
Llama 3.2 3B3B6.0 GB42 tok/s109K ctx
dense
Granite 3.1 8B8B9.3 GB42 tok/s71K ctx
state-space
Yi 1.5 9B9B9.5 GB33 tok/s4K ctx
dense
SmolLM3 3B3B6.3 GB42 tok/s96K ctx
dense
C
C Tier -- Tight Fit (163 models)
Runs with offloading or significant compromises.
Qwen 2.5 Math 7B7B7.6 GB43 tok/s4K ctx
dense
OpenChat 7B7B8.7 GB42 tok/s8K ctx
dense
DeepSeek R1 1.5B1.5B3.8 GB21 tok/s33K ctx
dense
Aya Expanse 8B8B9.3 GB37 tok/s8K ctx
dense
gemma 3 12b it
C5212B11.2 GB23 tok/s70K ctx
dense
TinyLlama 1.1B1.1B3.5 GB15 tok/s4K ctx
dense
DeepSeek R1 Distill Qwen 14B14B12.7 GB20 tok/s48K ctx
dense
Dolphin 2.9 8B8B9.3 GB37 tok/s33K ctx
dense
OpenHermes 2.5 7B7B8.7 GB42 tok/s8K ctx
dense
NousResearch Hermes 4 14B
C5214B12.7 GB20 tok/s48K ctx
dense
Vicuna 7B7B14.6 GB39 tok/s4K ctx
dense
Zephyr 7B Beta7B8.7 GB42 tok/s33K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C5214B12.7 GB20 tok/s48K ctx
dense
internlm JanusCoder 14B
C5214B12.7 GB20 tok/s48K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C5114B12.7 GB20 tok/s48K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C5110.7B10.3 GB26 tok/s89K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C5110.7B10.3 GB26 tok/s89K ctx
dense
Starling LM 7B7B8.7 GB42 tok/s8K ctx
dense
solar finalised finetuned Model 10.7B i1
C5110.7B10.3 GB26 tok/s89K ctx
dense
SOLAR 10.7B v1.0
C5110.7B10.3 GB26 tok/s89K ctx
dense
Qwen3.5 9B9B9.0 GB30 tok/s122K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C519B9.0 GB30 tok/s122K ctx
dense
Helply 10.2b chat i1
C5110.2B9.9 GB27 tok/s97K ctx
dense
Qwen3.5 9B
C519B9.0 GB30 tok/s122K ctx
dense
HelpingAI2.5 10B i1
C5110B9.8 GB27 tok/s101K ctx
dense
Yi Coder 9B Chat
C509B9.0 GB30 tok/s122K ctx
dense
glm 4 9b chat 1m
C509B9.0 GB30 tok/s122K ctx
dense
Qwen 2.5 1.5B1.5B3.8 GB21 tok/s131K ctx
dense
Nemotron Mini 4B4B6.9 GB56 tok/s4K ctx
dense
Neural Chat 7B7B8.7 GB42 tok/s8K ctx
dense
Yi 1.5 9B Chat
C509B9.0 GB30 tok/s122K ctx
dense
Meta Llama 3.1 8B Instruct
C508B8.3 GB34 tok/s147K ctx
dense
llava llama 3 8b v1 1
C508B8.3 GB34 tok/s147K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B8.3 GB34 tok/s147K ctx
dense
Yi 9B Coder i1
C509B9.0 GB30 tok/s122K ctx
dense
Meta Llama 3 8B Instruct
C508B8.3 GB34 tok/s147K ctx
dense
DeepSeek R1 Distill Llama 8B8B8.3 GB34 tok/s147K ctx
dense
Dolphin3.0 Llama3.1 8B
C508B8.3 GB34 tok/s147K ctx
dense
Llama 3 8B Instruct 32k v0.1
C508B8.3 GB34 tok/s147K ctx
dense
Meta Llama 3.1 8B Instruct
C508B8.3 GB34 tok/s147K ctx
dense
vntl llama3 8b v2
C508B8.3 GB34 tok/s147K ctx
dense
HelpingAI2 9B
C509B9.0 GB30 tok/s122K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C508B8.3 GB34 tok/s147K ctx
dense
HelpingAI2 9B i1
C509B9.0 GB30 tok/s122K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C508B8.3 GB34 tok/s147K ctx
dense
StarCoder2 15B15B14.5 GB17 tok/s16K ctx
dense
HelpingAI 9B 200k i1
C509B9.0 GB30 tok/s122K ctx
dense
Gemma 2 2B2B5.3 GB28 tok/s8K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C508B8.3 GB34 tok/s147K ctx
dense
HelpingAI 9B i1
C509B9.0 GB30 tok/s122K ctx
dense
Hermes 3 Llama 3.1 8B
C508B8.3 GB34 tok/s147K ctx
dense
Hermes 2 Pro Llama 3 8B
C508B8.3 GB34 tok/s147K ctx
dense
Dolphin3.0 Llama3.1 8B
C508B8.3 GB34 tok/s147K ctx
dense
granite 8b code instruct 4k
C508B8.3 GB34 tok/s147K ctx
dense
dolphin 2.9.4 llama3.1 8b
C508B8.3 GB34 tok/s147K ctx
dense
openchat 3.6 8b 20240522 IMat
C508B8.3 GB34 tok/s147K ctx
dense
Llama 2 7B Chat
C507B7.6 GB39 tok/s180K ctx
dense
falcon mamba 7b instruct Q4 K M
C507B7.6 GB45 tok/s180K ctx
dense
aya expanse 8b
C508B8.3 GB34 tok/s147K ctx
dense
Mistral 7B Instruct v0.2
C507B7.6 GB39 tok/s180K ctx
dense
dolphin v2 8b abliterated i1
C508B8.3 GB34 tok/s147K ctx
dense
EXAONE 3.5 7.8B Instruct
C507.8B8.2 GB35 tok/s153K ctx
dense
Mistral 7B Instruct v0.3
C507B7.6 GB39 tok/s180K ctx
dense
EXAONE 3.5 7.8B Instruct
C507.8B8.2 GB35 tok/s153K ctx
dense
aya expanse 8b orthogonal heretic i1
C508B8.3 GB34 tok/s147K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C497.8B8.2 GB35 tok/s153K ctx
dense
DeepSeek LLM 7B7B14.1 GB39 tok/s4K ctx
dense
exaone 3.0 7.8b it
C497.8B8.2 GB35 tok/s153K ctx
dense
aya expanse 8b orthogonal heretic
C498B8.3 GB34 tok/s147K ctx
dense
Mamba Codestral 7B v0.1
C497B7.6 GB45 tok/s180K ctx
dense
internlm2 5 1 8b chat i1
C498B8.3 GB34 tok/s147K ctx
dense
internlm3 8b instruct abliterated i1
C498B8.3 GB34 tok/s147K ctx
dense
Mistral 7B Instruct v0.3
C497B7.6 GB39 tok/s180K ctx
dense
Hermes 2 Pro Mistral 7B
C497B7.6 GB39 tok/s180K ctx
dense
zephyr 7B beta
C497B7.6 GB39 tok/s180K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C497B7.6 GB39 tok/s180K ctx
dense
Yi 1.5 6B6B7.1 GB50 tok/s4K ctx
dense
Falcon H1R 7B
C497B7.6 GB39 tok/s180K ctx
dense
zephyr 7B alpha
C497B7.6 GB39 tok/s180K ctx
dense
speechless zephyr code functionary 7b
C497B7.6 GB39 tok/s180K ctx
dense
StarCoder2 7B
C497B7.6 GB39 tok/s180K ctx
dense
japanese stablelm instruct gamma 7B
C497B7.6 GB39 tok/s180K ctx
dense
Falcon H1 7B Instruct
C497B7.6 GB39 tok/s180K ctx
dense
Falcon H1R 7B7B7.6 GB39 tok/s180K ctx
dense
Yi 1.5 6B Chat
C496B6.9 GB46 tok/s224K ctx
dense
stabilityai japanese stablelm base gamma 7b
C497B7.6 GB39 tok/s180K ctx
dense
baichuan2 7b chat
C497B7.6 GB39 tok/s180K ctx
dense
DiscoPOP zephyr 7b gemma
C497B7.6 GB39 tok/s180K ctx
dense
starcoder2 7b
C497B7.6 GB39 tok/s180K ctx
dense
stablelm 2 zephyr 1 6b
C496B6.9 GB46 tok/s224K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C497B7.6 GB39 tok/s180K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C497B7.6 GB39 tok/s180K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C497B7.6 GB39 tok/s180K ctx
dense
internlm2 math plus 7b IMat
C497B7.6 GB39 tok/s180K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C497B7.6 GB39 tok/s180K ctx
dense
internlm2 5 7b chat i1
C497B7.6 GB39 tok/s180K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C497B7.6 GB39 tok/s180K ctx
dense
MD Judge v0 2 internlm2 7b i1
C497B7.6 GB39 tok/s180K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C497B7.6 GB39 tok/s180K ctx
dense
jointpreferences mistral 7b sft helpful
C497B7.6 GB39 tok/s180K ctx
dense
zephyr 7b dpo full i1
C497B7.6 GB39 tok/s180K ctx
dense
blossom v3 baichuan2 7b i1
C497B7.6 GB39 tok/s180K ctx
dense
blossom v1 baichuan 7b i1
C497B7.6 GB39 tok/s180K ctx
dense
Yi 1.5 6B Chat
C496B6.9 GB46 tok/s224K ctx
dense
stablelm 2 1 6b chat imatrix
C496B6.9 GB46 tok/s224K ctx
dense
Phi 4 reasoning vision 15B
C4915B13.4 GB18 tok/s40K ctx
dense
StarCoder2 15B
C4915B13.4 GB18 tok/s40K ctx
dense
Gemma 3 1B1B3.5 GB14 tok/s33K ctx
dense
HelpingAI2 6B
C486B6.9 GB46 tok/s224K ctx
dense
StarCoder2 7B7B7.3 GB43 tok/s16K ctx
dense
starcoder2 15b instruct v0.1
C4815B13.4 GB18 tok/s40K ctx
dense
starcoder2 15b i1
C4815B13.4 GB18 tok/s40K ctx
dense
HelpingAI2 6B i1
C486B6.9 GB46 tok/s224K ctx
dense
starcoder2 15b instruct v0.1
C4815B13.4 GB18 tok/s40K ctx
dense
HelpingAI 15B i1
C4815B13.4 GB18 tok/s40K ctx
dense
HelpingAI2.5 5B i1
C485B6.1 GB55 tok/s285K ctx
dense
Qwen3.5 4B4B5.4 GB56 tok/s378K ctx
dense
gemma 3 4b it
C484B5.4 GB56 tok/s378K ctx
dense
gemma 3 4b it
C484B5.4 GB56 tok/s378K ctx
dense
Qwen 2.5 Coder 0.5B0.5B3.0 GB7 tok/s131K ctx
dense
Codestral RAG 19B Pruned i1
C4619B16.3 GB10 tok/s14K ctx
dense
Llama 3.2 3B Instruct
C463B5.0 GB42 tok/s516K ctx
dense
Codestral 22B22B18.4 GB8 tok/s4K ctx
dense
Qwen2.5 3B Instruct3B4.7 GB42 tok/s531K ctx
dense
Llama 3.2 3B Instruct
C463B4.7 GB42 tok/s531K ctx
dense
Ministral 3 3B Instruct 2512
C463B4.7 GB42 tok/s531K ctx
dense
HELVETE 3B
C463B4.7 GB42 tok/s531K ctx
dense
Qwen 3 0.6B0.6B3.7 GB8 tok/s33K ctx
dense
Hermes 3 Llama 3.2 3B
C453B4.7 GB42 tok/s531K ctx
dense
stablelm zephyr 3b
C453B4.7 GB42 tok/s531K ctx
dense
StarCoder2 3B
C453B4.7 GB42 tok/s531K ctx
dense
AI21 Jamba Reasoning 3B
C453B4.7 GB42 tok/s531K ctx
dense
Llama 3.2 1B1B3.6 GB14 tok/s128K ctx
dense
stablelm 3b 4e1t
C453B4.7 GB42 tok/s531K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C453B4.7 GB42 tok/s531K ctx
dense
ai21labs AI21 Jamba2 3B
C453B4.7 GB42 tok/s531K ctx
dense
HelpingAI 3B hindi i1
C453B4.7 GB42 tok/s531K ctx
dense
AI21 Jamba2 3B
C453B4.7 GB42 tok/s531K ctx
dense
HelpingAI 3B hindi
C453B4.7 GB42 tok/s531K ctx
dense
AI21 Jamba2 3B i1
C453B4.7 GB42 tok/s531K ctx
dense
StarCoder2 3B3B4.8 GB42 tok/s16K ctx
dense
gemma 2 2b it
C442B4.4 GB28 tok/s810K ctx
dense
EXAONE 3.5 2.4B Instruct
C442.4B4.2 GB34 tok/s685K ctx
dense
gemma 2b2B4.0 GB28 tok/s838K ctx
dense
Gemmasutra Mini 2B v1
C442B4.0 GB28 tok/s838K ctx
dense
gemma 2 2b it
C432B4.0 GB28 tok/s838K ctx
dense
Qwen 3.5 0.6B0.6B3.7 GB8 tok/s131K ctx
dense
Qwen2.5 1.5B Instruct1.5B3.6 GB21 tok/s1146K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B3.6 GB21 tok/s1146K ctx
dense
Yi Coder 1.5B Chat
C421.5B3.6 GB21 tok/s1146K ctx
dense
stablelm 2 zephyr 1.6b
C421.6B3.7 GB22 tok/s1069K ctx
dense
logos16v2 stablelm2 1.6b i1
C421.6B3.7 GB22 tok/s1069K ctx
dense
Yi Coder 1.5B
C421.5B3.6 GB21 tok/s1146K ctx
dense
Falcon H1 1.5B Instruct1.5B3.6 GB21 tok/s1146K ctx
dense
TinyLlama 1.1B Chat v1.0
C411.1B3.3 GB15 tok/s1592K ctx
dense
EXAONE 4.0 1.2B
C411.2B3.4 GB17 tok/s1453K ctx
dense
Llama 3.2 1B Instruct Q8 0
C411B3.4 GB14 tok/s1731K ctx
dense
TinyLlama 1.1B Chat v0.3
C411.1B3.3 GB15 tok/s1592K ctx
dense
TinyLlama 1.1B Chat v0.61.1B3.3 GB15 tok/s1592K ctx
dense
Qwen 2.5 0.5B0.5B3.0 GB7 tok/s131K ctx
dense
Llama 3.2 1B Instruct
C401B3.2 GB14 tok/s1760K ctx
dense
gemma 3 1b it
C401B3.2 GB14 tok/s1760K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
C401.1B3.3 GB15 tok/s1592K ctx
dense
F
F Tier -- Can't Run (128 models)
Too heavy for this hardware.
Falcon3 1B Instruct abliterated
D401B3.2 GB14 tok/s1760K ctx
dense
SmolVLM 500M Instruct
D380.5B3.0 GB7 tok/s2094K ctx
dense
embeddinggemma 300M
D360.3B2.8 GB4 tok/s2121K ctx
dense
internlm2 5 20b chat
D3620B17.0 GB9 tok/s9K ctx
dense
internlm2 limarp chat 20b
D3620B17.0 GB9 tok/s9K ctx
dense
internlm2 math plus 20b i1
D3620B17.0 GB9 tok/s9K ctx
dense
Codestral 22B v0.1
D3622B18.5 GB7 tok/s4K ctx
dense
Codestral 22B v0.1
D3622B18.5 GB7 tok/s4K ctx
dense
Codestral 21B Pruned i1
D3621B17.8 GB8 tok/s4K ctx
dense
Codestral 22B v0.1
D3522B18.5 GB7 tok/s4K ctx
dense
Codestral 22B v0.1 IMat
D3522B18.5 GB7 tok/s4K ctx
dense
Codestral 22B v0.1 i1
D3522B18.5 GB7 tok/s4K ctx
dense
granite embedding 107m multilingual
D340.11B2.7 GB2 tok/s2150K ctx
dense
Falcon H1 Tiny 90M Instruct
D340.09B2.7 GB2 tok/s2151K ctx
dense
Llama 3.1 70B70B50.1 GB2 tok/s4K ctx
dense
Llama 3.3 70B70B50.1 GB2 tok/s4K ctx
dense
Qwen 2.5 32B32B25.9 GB3 tok/s4K ctx
dense
Qwen 2.5 72B72B51.3 GB2 tok/s4K ctx
dense
Qwen 2.5 Coder 32B32B25.9 GB3 tok/s4K ctx
dense
Qwen 3 32B32B25.9 GB3 tok/s4K ctx
dense
Qwen 3 235B A22B235B148.7 GB2 tok/s4K ctx
moe
Qwen 3 30B A3B30.5B22.6 GB9 tok/s4K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B22.6 GB9 tok/s4K ctx
moe
Qwen3-Coder-Next80B52.8 GB2 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B298.2 GB2 tok/s4K ctx
moe
Qwen3-VL 30B A3B Instruct30B22.3 GB10 tok/s4K ctx
moe
Qwen 3.5 27B27B22.1 GB4 tok/s4K ctx
dense
Qwen 3.5 35B A3B35B25.3 GB7 tok/s4K ctx
moe
Qwen 3.6 35B A3B35B28.0 GB5 tok/s4K ctx
moe
Qwen 3.6 27B27B19.9 GB4 tok/s4K ctx
dense
Qwen 3.5 122B A10B122B79.4 GB2 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B247.5 GB2 tok/s4K ctx
moe
Gemma 2 27B27B30.2 GB2 tok/s4K ctx
dense
Gemma 3 27B27B30.2 GB2 tok/s4K ctx
dense
Mistral Small 24B24B19.6 GB6 tok/s4K ctx
dense
Mistral Small 3.2 24B24B19.6 GB6 tok/s4K ctx
vision
Mixtral 8x7B47B33.1 GB2 tok/s4K ctx
moe
Mixtral 8x22B141B91.9 GB2 tok/s4K ctx
moe
DeepSeek R1 Distill 32B32B25.9 GB3 tok/s4K ctx
dense
Granite 4.1 30B30B24.7 GB3 tok/s4K ctx
dense
Command R 35B35B26.3 GB2 tok/s4K ctx
dense
Falcon 40B Instruct40B33.1 GB2 tok/s4K ctx
dense
WizardLM 13B13B22.6 GB8 tok/s7K ctx
dense
StarCoder 15B15B27.9 GB4 tok/s4K ctx
dense
CodeLlama 13B Instruct13B22.6 GB8 tok/s7K ctx
dense
MPT-30B-Instruct30B47.5 GB2 tok/s4K ctx
dense
Devstral Small 1.124B19.6 GB6 tok/s4K ctx
dense
Devstral Small 2 24B Instruct24B19.6 GB6 tok/s4K ctx
dense
Devstral 2 123B Instruct123B82.9 GB2 tok/s4K ctx
dense
Mistral Small 4 119B119B80.5 GB2 tok/s4K ctx
moe
Leanstral 119B A6B119B83.9 GB2 tok/s4K ctx
moe
Mistral Large 3675B419.6 GB2 tok/s4K ctx
moe
Magistral Small 250724B19.6 GB6 tok/s4K ctx
dense
GLM-5744B475.4 GB2 tok/s4K ctx
moe
GLM-5.1754B481.5 GB2 tok/s4K ctx
moe
Kimi K2.51000B619.9 GB2 tok/s4K ctx
moe
Kimi K2.61000B619.9 GB2 tok/s4K ctx
moe
Kimi Linear 48B A3B48B32.7 GB2 tok/s4K ctx
linear
Nous Hermes 1.09B20.2 GB14 tok/s10K ctx
dense
Nous Dolphin 13B13B24.1 GB6 tok/s5K ctx
dense
Baichuan 13B13B24.1 GB6 tok/s5K ctx
dense
Cerebras-GPT 13B13B21.6 GB7 tok/s7K ctx
dense
StableLM 2 12B12B23.3 GB6 tok/s4K ctx
dense
InternLM 20B20B35.2 GB2 tok/s4K ctx
dense
Yi 34B Chat34B26.9 GB2 tok/s4K ctx
dense
DeepSeek V3 671B671B471.4 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B471.4 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B866.4 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B161.8 GB2 tok/s4K ctx
moe
DeepSeek R1 671B671B471.4 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B205.1 GB2 tok/s4K ctx
moe
DeepSeek V2.5 236B236B205.1 GB2 tok/s4K ctx
moe
DeepSeek LLM 67B67B49.2 GB2 tok/s4K ctx
dense
Llama 4 Scout 17B 16E109B71.9 GB2 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B249.4 GB2 tok/s4K ctx
moe
Command R+ 104B104B69.4 GB2 tok/s4K ctx
dense
Command A 111B111B74.1 GB2 tok/s4K ctx
dense
Yi 1.5 34B34B26.9 GB2 tok/s4K ctx
dense
Pixtral Large 124B124B83.5 GB2 tok/s4K ctx
dense
Qwen 2.5 VL 72B72B51.3 GB2 tok/s4K ctx
dense
Qwen 2.5 Math 72B72B51.3 GB2 tok/s4K ctx
dense
Nemotron 70B70B50.1 GB2 tok/s4K ctx
dense
LLaVA 1.6 13B13B22.6 GB8 tok/s4K ctx
dense
Vicuna 13B13B22.6 GB8 tok/s4K ctx
dense
OLMo 2 32B32B25.9 GB3 tok/s4K ctx
dense
Nemotron 3 Nano 30B30B23.2 GB3 tok/s4K ctx
dense
EXAONE 4.0 32B32B25.9 GB3 tok/s4K ctx
dense
LFM2 24B24B19.6 GB6 tok/s4K ctx
dense
Llama 3.1 405B405B257.2 GB2 tok/s4K ctx
dense
DeepSeek R1 Distill 70B70B50.1 GB2 tok/s4K ctx
dense
Granite Code 34B34B26.9 GB2 tok/s4K ctx
dense
GPT-OSS 120B117B78.8 GB2 tok/s4K ctx
dense
Aya Expanse 32B32B24.5 GB3 tok/s4K ctx
dense
Mistral Small 3.1 24B24B19.6 GB6 tok/s4K ctx
dense
Gemma 4 26B A4B25.2B21.5 GB11 tok/s4K ctx
moe
Gemma 4 31B30.7B35.9 GB2 tok/s4K ctx
dense
DeepSeek V3.2671B412.3 GB2 tok/s4K ctx
moe
Nemotron Cascade 2 30B A3B30B23.7 GB8 tok/s4K ctx
moe
MiniMax M2.7
F0230B146.6 GB2 tok/s4K ctx
moe
Solar Open 100B
F0100B75.2 GB2 tok/s4K ctx
dense
Baichuan M2 32B Q4 K M
F032B25.8 GB2 tok/s4K ctx
dense
baichuan inc Baichuan M2 32B
F032B25.8 GB2 tok/s4K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
F024B20.0 GB5 tok/s4K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
F024B20.0 GB5 tok/s4K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
F048B37.4 GB2 tok/s4K ctx
dense
EXAONE 4.0 32B
F032B25.8 GB2 tok/s4K ctx
dense
K EXAONE 236B A23B
F0236B174.1 GB2 tok/s4K ctx
dense
Qwen3.5 35B A3B
F035B28.0 GB2 tok/s4K ctx
dense
gemma 3 27b it
F027B22.1 GB4 tok/s4K ctx
dense
Llama 3.3 70B Instruct
F070B53.4 GB2 tok/s4K ctx
dense
Mistral Small 24B Instruct 2501
F024B20.0 GB5 tok/s4K ctx
dense
mistral small 3.1 24b instruct 2503 hf
F024B20.0 GB5 tok/s4K ctx
dense
aya expanse 32b heretic MPOA i1
F032B25.8 GB2 tok/s4K ctx
dense
Baichuan M3 235B
F0235B173.4 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B173.4 GB2 tok/s4K ctx
dense
BaichuanMed OCR 72B i1
F072B54.9 GB2 tok/s4K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
F024B20.0 GB5 tok/s4K ctx
dense
Solar Open 100B i1
F0100B75.2 GB2 tok/s4K ctx
dense
Solar Open 69B REAP i1
F069B52.7 GB2 tok/s4K ctx
dense
Hermes 4.3 36B
F036B28.7 GB2 tok/s4K ctx
dense
stabilityai japanese stablelm instruct beta 70b
F070B53.4 GB2 tok/s4K ctx
dense
gemma 3 27b it27B22.1 GB4 tok/s4K ctx
dense
Mistral Small 3.2 24B Instruct 250624B20.0 GB5 tok/s4K ctx
dense
Qwen3.5 122B A10B122B76.6 GB2 tok/s4K ctx
dense
Qwen3.5 27B27B22.1 GB4 tok/s4K ctx
dense
Qwen3.5 35B A3B35B28.0 GB2 tok/s4K ctx
dense
Qwen3.5 397B A17B397B291.2 GB2 tok/s4K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
F024B20.0 GB5 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs