AI Compatibility Tier List
RTX 4000 Ada 20GB
374 AI models ranked by how well they run on your 20 GB VRAM hardware. Best match: Qwen 3 14B (score: 94).
374
Models tested
279
Can run
20 GB
VRAM
36
Best tok/s
Tier Distribution
S 13
A 54
B 36
C 176
F 95
S: 13A: 54B: 36C: 176F: 95
S
S Tier -- Perfect Fit (13 models)
These models run flawlessly with headroom to spare.
S94
14B13.9 GB36 tok/s56K ctx
dense
S93
14.7B14.9 GB34 tok/s33K ctx
dense
S93
9B10.6 GB55 tok/s85K ctx
dense
S91
8B10.0 GB62 tok/s89K ctx
dense
S91
21B18.2 GB54 tok/s28K ctx
moe
S91
24B20.0 GB21 tok/s16K ctx
dense
S90
24B20.0 GB21 tok/s16K ctx
dense
S89
27B20.3 GB10 tok/s10K ctx
+1dense
S89
14B13.9 GB35 tok/s56K ctx
multimodal
S89
24B20.0 GB21 tok/s16K ctx
dense
S88
4B7.5 GB56 tok/s107K ctx
dense
S86
8B9.7 GB62 tok/s100K ctx
dense
S85
14B14.5 GB35 tok/s16K ctx
dense
A
A Tier -- Great Fit (54 models)
Excellent performance with comfortable memory margins.
A85
14B14.4 GB36 tok/s47K ctx
dense
A85
16B16.0 GB69 tok/s36K ctx
moe
A84
3.8B6.7 GB53 tok/s131K ctx
dense
A83
8B9.7 GB62 tok/s8K ctx
dense
A83
22B18.8 GB20 tok/s24K ctx
dense
A83
19B16.9 GB26 tok/s8K ctx
dense
A83
24B20.0 GB21 tok/s16K ctx
vision
A83
12B15.1 GB31 tok/s32K ctx
dense
A83
30.5B23.0 GB24 tok/s4K ctx
moe
A82
24B20.0 GB21 tok/s16K ctx
dense
A82
30B22.7 GB25 tok/s4K ctx
moe
A82
8B10.0 GB62 tok/s89K ctx
multimodal
A82
8B9.7 GB62 tok/s2K ctx
dense
A81
9B10.8 GB55 tok/s76K ctx
dense
A81
24B20.0 GB21 tok/s16K ctx
dense
A80
7B9.1 GB71 tok/s8K ctx
dense
A80
24B20.0 GB21 tok/s16K ctx
dense
A80
13B13.3 GB38 tok/s33K ctx
dense
A80
30.5B23.0 GB24 tok/s4K ctx
moe
A80
4B7.5 GB56 tok/s33K ctx
dense
A80
27B22.5 GB11 tok/s4K ctx
dense
A79
20B18.3 GB25 tok/s8K ctx
dense
A79
7B8.0 GB71 tok/s33K ctx
dense
A79
7B14.5 GB66 tok/s8K ctx
dense
A79
7B15.0 GB66 tok/s16K ctx
dense
A79
7B9.1 GB71 tok/s8K ctx
dense
A78
14B14.4 GB36 tok/s33K ctx
dense
A78
9B9.0 GB56 tok/s131K ctx
dense
A77
0.57B6.0 GB8 tok/s8K ctx
dense
A77
8B9.1 GB47 tok/s128K ctx
dense
A76
12B12.7 GB41 tok/s64K ctx
dense
A76
7B15.0 GB66 tok/s8K ctx
dense
A76
7B15.0 GB66 tok/s8K ctx
dense
A76
8B9.7 GB62 tok/s8K ctx
dense
A76
7B9.1 GB71 tok/s8K ctx
dense
A76
8B10.2 GB62 tok/s80K ctx
dense
A76
0.57B5.2 GB8 tok/s8K ctx
dense
A76
7B8.0 GB71 tok/s131K ctx
dense
A75
25.2B21.9 GB28 tok/s8K ctx
moe
A75
7B7.7 GB76 tok/s262K ctx
state-space
A74
7B15.0 GB66 tok/s4K ctx
dense
A72
7B10.1 GB71 tok/s8K ctx
dense
A72
3B6.9 GB42 tok/s111K ctx
dense
A72
0.34B5.1 GB5 tok/s1K ctx
dense
A72
8B9.7 GB62 tok/s100K ctx
dense
A72
7B15.0 GB66 tok/s8K ctx
dense
A72
7B15.0 GB66 tok/s8K ctx
dense
A71
9B20.6 GB36 tok/s15K ctx
dense
A71
7B9.1 GB71 tok/s4K ctx
dense
A71
7B9.1 GB71 tok/s4K ctx
dense
A71
0.34B5.1 GB5 tok/s1K ctx
dense
A71
9B9.0 GB56 tok/s128K ctx
dense
A71
5.1B6.5 GB71 tok/s128K ctx
dense
A71
30B23.6 GB9 tok/s4K ctx
dense
B
B Tier -- Good Fit (36 models)
Solid performance, may need quantization or reduced context.
B70
7B8.0 GB71 tok/s131K ctx
dense
B70
0.34B5.1 GB5 tok/s1K ctx
dense
B69
4B7.4 GB56 tok/s113K ctx
dense
B69
3.8B11.1 GB53 tok/s40K ctx
dense
B68
9B13.5 GB41 tok/s8K ctx
dense
B68
4B6.8 GB56 tok/s128K ctx
dense
B68
14B14.4 GB36 tok/s47K ctx
dense
B68
4B11.2 GB56 tok/s40K ctx
dense
B68
3B5.5 GB42 tok/s262K ctx
multimodal
B68
8B9.7 GB62 tok/s33K ctx
dense
B67
7B9.1 GB71 tok/s4K ctx
dense
B67
0.14B3.7 GB2 tok/s8K ctx
dense
B67
7B8.0 GB71 tok/s33K ctx
dense
B67
2B5.8 GB28 tok/s131K ctx
dense
B66
11B11.6 GB45 tok/s16K ctx
vision
B66
3B6.9 GB42 tok/s111K ctx
dense
B65
12B12.7 GB41 tok/s64K ctx
dense
B65
7B7.3 GB72 tok/s8K ctx
dense
B65
14B14.5 GB35 tok/s45K ctx
dense
B64
13B23.0 GB20 tok/s12K ctx
dense
B63
1.7B5.6 GB24 tok/s33K ctx
dense
B63
3B7.2 GB42 tok/s8K ctx
dense
B63
9B9.9 GB56 tok/s127K ctx
dense
B63
7B9.1 GB71 tok/s8K ctx
dense
B63
3B6.0 GB42 tok/s131K ctx
dense
B62
13B23.0 GB20 tok/s4K ctx
dense
B61
0.02B3.2 GB2 tok/s0K ctx
dense
B61
8B10.0 GB62 tok/s89K ctx
dense
B60
1.5B4.2 GB21 tok/s33K ctx
dense
B60
13B23.0 GB20 tok/s8K ctx
dense
B60
13B23.0 GB20 tok/s4K ctx
dense
B59
22B18.8 GB21 tok/s24K ctx
dense
B59
3B6.4 GB42 tok/s128K ctx
dense
B56
7B15.0 GB66 tok/s4K ctx
dense
B56
8B9.7 GB71 tok/s100K ctx
state-space
B55
9B9.9 GB56 tok/s4K ctx
dense
C
C Tier -- Tight Fit (176 models)
Runs with offloading or significant compromises.
C55
13B22.0 GB19 tok/s13K ctx
dense
C55
15B14.9 GB29 tok/s16K ctx
dense
C54
3B6.7 GB42 tok/s125K ctx
dense
C54
7B14.5 GB66 tok/s4K ctx
dense
C54
7B8.0 GB71 tok/s4K ctx
dense
Phi 4 reasoning vision 15B
C5315B13.8 GB31 tok/s72K ctx
dense
StarCoder2 15B
C5315B13.8 GB31 tok/s72K ctx
dense
starcoder2 15b instruct v0.1
C5315B13.8 GB31 tok/s72K ctx
dense
C53
7B9.1 GB71 tok/s8K ctx
dense
starcoder2 15b i1
C5315B13.8 GB31 tok/s72K ctx
dense
starcoder2 15b instruct v0.1
C5315B13.8 GB31 tok/s72K ctx
dense
C53
14B13.1 GB33 tok/s83K ctx
dense
HelpingAI 15B i1
C5315B13.8 GB31 tok/s72K ctx
dense
C53
8B9.7 GB62 tok/s8K ctx
dense
NousResearch Hermes 4 14B
C5214B13.1 GB33 tok/s83K ctx
dense
C52
1.5B4.2 GB21 tok/s33K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C5214B13.1 GB33 tok/s83K ctx
dense
internlm JanusCoder 14B
C5214B13.1 GB33 tok/s83K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C5214B13.1 GB33 tok/s83K ctx
dense
C52
1.1B3.9 GB15 tok/s4K ctx
dense
C52
8B9.7 GB62 tok/s33K ctx
dense
gemma 3 12b it
C5212B11.6 GB38 tok/s111K ctx
dense
C52
7B9.1 GB71 tok/s8K ctx
dense
C51
7B9.1 GB71 tok/s33K ctx
dense
C51
7B9.1 GB71 tok/s8K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C5110.7B10.7 GB43 tok/s135K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C5110.7B10.7 GB43 tok/s135K ctx
dense
solar finalised finetuned Model 10.7B i1
C5110.7B10.7 GB43 tok/s135K ctx
dense
SOLAR 10.7B v1.0
C5010.7B10.7 GB43 tok/s135K ctx
dense
C50
9B9.4 GB51 tok/s176K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C509B9.4 GB51 tok/s176K ctx
dense
Qwen3.5 9B
C509B9.4 GB51 tok/s176K ctx
dense
Yi Coder 9B Chat
C509B9.4 GB51 tok/s176K ctx
dense
glm 4 9b chat 1m
C509B9.4 GB51 tok/s176K ctx
dense
Helply 10.2b chat i1
C5010.2B10.3 GB45 tok/s146K ctx
dense
HelpingAI2.5 10B i1
C5010B10.2 GB46 tok/s150K ctx
dense
C50
7B9.1 GB71 tok/s8K ctx
dense
Meta Llama 3.1 8B Instruct
C508B8.7 GB58 tok/s209K ctx
dense
llava llama 3 8b v1 1
C508B8.7 GB58 tok/s209K ctx
dense
C50
8B8.7 GB58 tok/s209K ctx
dense
Yi 1.5 9B Chat
C509B9.4 GB51 tok/s176K ctx
dense
Meta Llama 3 8B Instruct
C508B8.7 GB58 tok/s209K ctx
dense
C50
8B8.7 GB58 tok/s209K ctx
dense
Dolphin3.0 Llama3.1 8B
C508B8.7 GB58 tok/s209K ctx
dense
Llama 3 8B Instruct 32k v0.1
C508B8.7 GB58 tok/s209K ctx
dense
Meta Llama 3.1 8B Instruct
C508B8.7 GB58 tok/s209K ctx
dense
vntl llama3 8b v2
C508B8.7 GB58 tok/s209K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C508B8.7 GB58 tok/s209K ctx
dense
C50
1.5B4.2 GB21 tok/s131K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C508B8.7 GB58 tok/s209K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C508B8.7 GB58 tok/s209K ctx
dense
Yi 9B Coder i1
C509B9.4 GB51 tok/s176K ctx
dense
Hermes 3 Llama 3.1 8B
C508B8.7 GB58 tok/s209K ctx
dense
Llama 2 7B Chat
C507B8.0 GB66 tok/s250K ctx
dense
falcon mamba 7b instruct Q4 K M
C507B8.0 GB76 tok/s250K ctx
dense
Mistral 7B Instruct v0.2
C507B8.0 GB66 tok/s250K ctx
dense
HelpingAI2 9B
C509B9.4 GB51 tok/s176K ctx
dense
Hermes 2 Pro Llama 3 8B
C508B8.7 GB58 tok/s209K ctx
dense
Dolphin3.0 Llama3.1 8B
C508B8.7 GB58 tok/s209K ctx
dense
Mistral 7B Instruct v0.3
C507B8.0 GB66 tok/s250K ctx
dense
granite 8b code instruct 4k
C508B8.7 GB58 tok/s209K ctx
dense
dolphin 2.9.4 llama3.1 8b
C508B8.7 GB58 tok/s209K ctx
dense
HelpingAI2 9B i1
C509B9.4 GB51 tok/s176K ctx
dense
HelpingAI 9B 200k i1
C509B9.4 GB51 tok/s176K ctx
dense
HelpingAI 9B i1
C509B9.4 GB51 tok/s176K ctx
dense
openchat 3.6 8b 20240522 IMat
C508B8.7 GB58 tok/s209K ctx
dense
aya expanse 8b
C498B8.7 GB58 tok/s209K ctx
dense
EXAONE 3.5 7.8B Instruct
C497.8B8.6 GB59 tok/s216K ctx
dense
Mamba Codestral 7B v0.1
C497B8.0 GB76 tok/s250K ctx
dense
dolphin v2 8b abliterated i1
C498B8.7 GB58 tok/s209K ctx
dense
EXAONE 3.5 7.8B Instruct
C497.8B8.6 GB59 tok/s216K ctx
dense
C49
6B7.5 GB83 tok/s4K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C497.8B8.6 GB59 tok/s216K ctx
dense
aya expanse 8b orthogonal heretic i1
C498B8.7 GB58 tok/s209K ctx
dense
Mistral 7B Instruct v0.3
C497B8.0 GB66 tok/s250K ctx
dense
exaone 3.0 7.8b it
C497.8B8.6 GB59 tok/s216K ctx
dense
Hermes 2 Pro Mistral 7B
C497B8.0 GB66 tok/s250K ctx
dense
zephyr 7B beta
C497B8.0 GB66 tok/s250K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C497B8.0 GB66 tok/s250K ctx
dense
aya expanse 8b orthogonal heretic
C498B8.7 GB58 tok/s209K ctx
dense
Falcon H1R 7B
C497B8.0 GB66 tok/s250K ctx
dense
zephyr 7B alpha
C497B8.0 GB66 tok/s250K ctx
dense
speechless zephyr code functionary 7b
C497B8.0 GB66 tok/s250K ctx
dense
Yi 1.5 6B Chat
C496B7.3 GB77 tok/s306K ctx
dense
internlm2 5 1 8b chat i1
C498B8.7 GB58 tok/s209K ctx
dense
internlm3 8b instruct abliterated i1
C498B8.7 GB58 tok/s209K ctx
dense
StarCoder2 7B
C497B8.0 GB66 tok/s250K ctx
dense
japanese stablelm instruct gamma 7B
C497B8.0 GB66 tok/s250K ctx
dense
stablelm 2 zephyr 1 6b
C496B7.3 GB77 tok/s306K ctx
dense
Falcon H1 7B Instruct
C497B8.0 GB66 tok/s250K ctx
dense
C49
7B8.0 GB66 tok/s250K ctx
dense
Codestral 22B v0.1
C4922B18.9 GB21 tok/s23K ctx
dense
Codestral RAG 19B Pruned i1
C4919B16.7 GB24 tok/s40K ctx
dense
Codestral 22B v0.1
C4922B18.9 GB21 tok/s23K ctx
dense
internlm2 5 20b chat
C4920B17.4 GB23 tok/s33K ctx
dense
stabilityai japanese stablelm base gamma 7b
C497B8.0 GB66 tok/s250K ctx
dense
baichuan2 7b chat
C497B8.0 GB66 tok/s250K ctx
dense
internlm2 limarp chat 20b
C4920B17.4 GB23 tok/s33K ctx
dense
DiscoPOP zephyr 7b gemma
C497B8.0 GB66 tok/s250K ctx
dense
Yi 1.5 6B Chat
C496B7.3 GB77 tok/s306K ctx
dense
stablelm 2 1 6b chat imatrix
C496B7.3 GB77 tok/s306K ctx
dense
starcoder2 7b
C497B8.0 GB66 tok/s250K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C497B8.0 GB66 tok/s250K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C497B8.0 GB66 tok/s250K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C497B8.0 GB66 tok/s250K ctx
dense
internlm2 math plus 20b i1
C4920B17.4 GB23 tok/s33K ctx
dense
internlm2 math plus 7b IMat
C497B8.0 GB66 tok/s250K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C497B8.0 GB66 tok/s250K ctx
dense
C49
4B7.3 GB56 tok/s4K ctx
dense
internlm2 5 7b chat i1
C497B8.0 GB66 tok/s250K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C497B8.0 GB66 tok/s250K ctx
dense
MD Judge v0 2 internlm2 7b i1
C497B8.0 GB66 tok/s250K ctx
dense
Codestral 22B v0.1
C4922B18.9 GB21 tok/s23K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C497B8.0 GB66 tok/s250K ctx
dense
C49
2B5.7 GB28 tok/s8K ctx
dense
jointpreferences mistral 7b sft helpful
C497B8.0 GB66 tok/s250K ctx
dense
zephyr 7b dpo full i1
C497B8.0 GB66 tok/s250K ctx
dense
blossom v3 baichuan2 7b i1
C497B8.0 GB66 tok/s250K ctx
dense
blossom v1 baichuan 7b i1
C497B8.0 GB66 tok/s250K ctx
dense
Codestral 21B Pruned i1
C4921B18.2 GB22 tok/s28K ctx
dense
Codestral 22B v0.1 IMat
C4922B18.9 GB21 tok/s23K ctx
dense
HelpingAI2 6B
C496B7.3 GB77 tok/s306K ctx
dense
HelpingAI2 6B i1
C496B7.3 GB77 tok/s306K ctx
dense
Codestral 22B v0.1 i1
C4922B18.9 GB21 tok/s23K ctx
dense
C49
7B7.7 GB72 tok/s16K ctx
dense
C48
1B3.9 GB14 tok/s33K ctx
dense
C48
24B20.4 GB14 tok/s14K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4824B20.4 GB14 tok/s14K ctx
dense
Mistral Small 24B Instruct 2501
C4824B20.4 GB14 tok/s14K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4824B20.4 GB14 tok/s14K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4824B20.4 GB14 tok/s14K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4824B20.4 GB14 tok/s14K ctx
dense
HelpingAI2.5 5B i1
C475B6.5 GB70 tok/s384K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4724B20.4 GB14 tok/s14K ctx
dense
C47
4B5.8 GB56 tok/s500K ctx
dense
gemma 3 4b it
C474B5.8 GB56 tok/s500K ctx
dense
gemma 3 4b it
C474B5.8 GB56 tok/s500K ctx
dense
C46
0.5B3.4 GB7 tok/s131K ctx
dense
Llama 3.2 3B Instruct
C453B5.4 GB42 tok/s680K ctx
dense
C45
3B5.1 GB42 tok/s695K ctx
dense
C45
0.6B4.1 GB8 tok/s33K ctx
dense
Llama 3.2 3B Instruct
C453B5.1 GB42 tok/s695K ctx
dense
Ministral 3 3B Instruct 2512
C453B5.1 GB42 tok/s695K ctx
dense
C45
1B4.0 GB14 tok/s128K ctx
dense
HELVETE 3B
C453B5.1 GB42 tok/s695K ctx
dense
Hermes 3 Llama 3.2 3B
C453B5.1 GB42 tok/s695K ctx
dense
stablelm zephyr 3b
C453B5.1 GB42 tok/s695K ctx
dense
StarCoder2 3B
C453B5.1 GB42 tok/s695K ctx
dense
AI21 Jamba Reasoning 3B
C453B5.1 GB42 tok/s695K ctx
dense
stablelm 3b 4e1t
C443B5.1 GB42 tok/s695K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C443B5.1 GB42 tok/s695K ctx
dense
ai21labs AI21 Jamba2 3B
C443B5.1 GB42 tok/s695K ctx
dense
HelpingAI 3B hindi i1
C443B5.1 GB42 tok/s695K ctx
dense
AI21 Jamba2 3B
C443B5.1 GB42 tok/s695K ctx
dense
HelpingAI 3B hindi
C443B5.1 GB42 tok/s695K ctx
dense
AI21 Jamba2 3B i1
C443B5.1 GB42 tok/s695K ctx
dense
gemma 2 2b it
C442B4.8 GB28 tok/s1055K ctx
dense
C43
3B5.2 GB42 tok/s16K ctx
dense
EXAONE 3.5 2.4B Instruct
C432.4B4.6 GB34 tok/s890K ctx
dense
C43
2B4.4 GB28 tok/s1084K ctx
dense
Gemmasutra Mini 2B v1
C432B4.4 GB28 tok/s1084K ctx
dense
gemma 2 2b it
C432B4.4 GB28 tok/s1084K ctx
dense
C43
0.6B4.1 GB8 tok/s131K ctx
dense
C42
1.5B4.0 GB21 tok/s1473K ctx
dense
C42
1.5B4.0 GB21 tok/s1473K ctx
dense
Yi Coder 1.5B Chat
C421.5B4.0 GB21 tok/s1473K ctx
dense
stablelm 2 zephyr 1.6b
C411.6B4.1 GB22 tok/s1376K ctx
dense
logos16v2 stablelm2 1.6b i1
C411.6B4.1 GB22 tok/s1376K ctx
dense
Yi Coder 1.5B
C411.5B4.0 GB21 tok/s1473K ctx
dense
C41
1.5B4.0 GB21 tok/s1473K ctx
dense
TinyLlama 1.1B Chat v1.0
C401.1B3.7 GB15 tok/s2039K ctx
dense
EXAONE 4.0 1.2B
C401.2B3.8 GB17 tok/s1862K ctx
dense
Llama 3.2 1B Instruct Q8 0
C401B3.8 GB14 tok/s2223K ctx
dense
TinyLlama 1.1B Chat v0.3
C401.1B3.7 GB15 tok/s2039K ctx
dense
C40
1.1B3.7 GB15 tok/s2039K ctx
dense
C40
0.5B3.4 GB7 tok/s131K ctx
dense
F
F Tier -- Can't Run (95 models)
Too heavy for this hardware.
Llama 3.2 1B Instruct
D401B3.6 GB14 tok/s2251K ctx
dense
gemma 3 1b it
D401B3.6 GB14 tok/s2251K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D401.1B3.7 GB15 tok/s2039K ctx
dense
Falcon3 1B Instruct abliterated
D391B3.6 GB14 tok/s2251K ctx
dense
SmolVLM 500M Instruct
D380.5B3.4 GB7 tok/s2670K ctx
dense
D37
27B22.5 GB10 tok/s4K ctx
dense
D37
27B22.5 GB10 tok/s4K ctx
dense
gemma 3 27b it
D3727B22.5 GB10 tok/s4K ctx
dense
embeddinggemma 300M
D360.3B3.2 GB4 tok/s2697K ctx
dense
granite embedding 107m multilingual
D340.11B3.1 GB2 tok/s2726K ctx
dense
Falcon H1 Tiny 90M Instruct
D340.09B3.1 GB2 tok/s2727K ctx
dense
D32
12B23.7 GB16 tok/s4K ctx
dense
F0
70B50.5 GB2 tok/s4K ctx
dense
F0
70B50.5 GB2 tok/s4K ctx
dense
F0
32B26.3 GB7 tok/s4K ctx
dense
F0
72B51.7 GB2 tok/s4K ctx
dense
F0
32B26.3 GB7 tok/s4K ctx
dense
F0
32B26.3 GB7 tok/s4K ctx
dense
F0
235B149.1 GB2 tok/s4K ctx
moe
F0
80B53.2 GB3 tok/s4K ctx
moe
F0
480B298.6 GB2 tok/s4K ctx
moe
F0
35B25.7 GB17 tok/s4K ctx
moe
F0
35B28.4 GB13 tok/s4K ctx
+1moe
F0
122B79.8 GB2 tok/s4K ctx
moe
F0
397B247.9 GB2 tok/s4K ctx
moe
F0
27B30.6 GB4 tok/s4K ctx
dense
F0
27B30.6 GB4 tok/s4K ctx
dense
F0
47B33.5 GB5 tok/s4K ctx
moe
F0
141B92.3 GB2 tok/s4K ctx
moe
F0
32B26.3 GB7 tok/s4K ctx
dense
F0
30B25.1 GB8 tok/s4K ctx
dense
F0
35B26.7 GB6 tok/s4K ctx
dense
F0
40B33.5 GB3 tok/s4K ctx
dense
F0
15B28.3 GB10 tok/s7K ctx
dense
F0
30B47.9 GB2 tok/s4K ctx
dense
F0
123B83.3 GB2 tok/s4K ctx
dense
F0
119B80.9 GB2 tok/s4K ctx
moe
F0
119B84.3 GB2 tok/s4K ctx
moe
F0
675B420.0 GB2 tok/s4K ctx
+1moe
F0
744B475.8 GB2 tok/s4K ctx
moe
F0
754B481.9 GB2 tok/s4K ctx
moe
F0
1000B620.3 GB2 tok/s4K ctx
moe
F0
1000B620.3 GB2 tok/s4K ctx
+1moe
F0
48B33.1 GB3 tok/s4K ctx
linear
F0
13B24.5 GB15 tok/s10K ctx
dense
F0
13B24.5 GB15 tok/s8K ctx
dense
F0
20B35.6 GB4 tok/s4K ctx
dense
F0
34B27.3 GB6 tok/s4K ctx
dense
F0
671B471.8 GB2 tok/s4K ctx
moe
F0
671B471.8 GB2 tok/s4K ctx
moe
F0
1600B866.8 GB2 tok/s4K ctx
moe
F0
284B162.2 GB2 tok/s4K ctx
moe
F0
671B471.8 GB2 tok/s4K ctx
moe
F0
236B205.5 GB2 tok/s4K ctx
moe
F0
236B205.5 GB2 tok/s4K ctx
moe
F0
67B49.6 GB2 tok/s4K ctx
dense
F0
109B72.3 GB2 tok/s4K ctx
moe
F0
400B249.8 GB2 tok/s4K ctx
moe
F0
104B69.8 GB2 tok/s4K ctx
dense
F0
111B74.5 GB2 tok/s4K ctx
dense
F0
34B27.3 GB6 tok/s4K ctx
dense
F0
124B83.9 GB2 tok/s4K ctx
dense
F0
72B51.7 GB2 tok/s4K ctx
dense
F0
72B51.7 GB2 tok/s4K ctx
dense
F0
70B50.5 GB2 tok/s4K ctx
dense
F0
32B26.3 GB7 tok/s4K ctx
dense
F0
32B26.3 GB7 tok/s4K ctx
dense
F0
405B257.6 GB2 tok/s4K ctx
dense
F0
70B50.5 GB2 tok/s4K ctx
dense
F0
34B27.3 GB6 tok/s4K ctx
dense
F0
117B79.2 GB2 tok/s4K ctx
dense
F0
32B24.9 GB7 tok/s4K ctx
dense
F0
30.7B36.3 GB3 tok/s4K ctx
dense
F0
671B412.7 GB2 tok/s4K ctx
moe
F0
30B24.1 GB22 tok/s4K ctx
moe
MiniMax M2.7
F0230B147.0 GB2 tok/s4K ctx
moe
Solar Open 100B
F0100B75.6 GB2 tok/s4K ctx
dense
Baichuan M2 32B Q4 K M
F032B26.2 GB6 tok/s4K ctx
dense
baichuan inc Baichuan M2 32B
F032B26.2 GB6 tok/s4K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
F048B37.8 GB2 tok/s4K ctx
dense
EXAONE 4.0 32B
F032B26.2 GB6 tok/s4K ctx
dense
K EXAONE 236B A23B
F0236B174.5 GB2 tok/s4K ctx
dense
Qwen3.5 35B A3B
F035B28.4 GB5 tok/s4K ctx
dense
Llama 3.3 70B Instruct
F070B53.8 GB2 tok/s4K ctx
dense
aya expanse 32b heretic MPOA i1
F032B26.2 GB6 tok/s4K ctx
dense
Baichuan M3 235B
F0235B173.8 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B173.8 GB2 tok/s4K ctx
dense
BaichuanMed OCR 72B i1
F072B55.3 GB2 tok/s4K ctx
dense
Solar Open 100B i1
F0100B75.6 GB2 tok/s4K ctx
dense
Solar Open 69B REAP i1
F069B53.1 GB2 tok/s4K ctx
dense
Hermes 4.3 36B
F036B29.1 GB4 tok/s4K ctx
dense
stabilityai japanese stablelm instruct beta 70b
F070B53.8 GB2 tok/s4K ctx
dense
F0
122B77.0 GB2 tok/s4K ctx
dense
F0
35B28.4 GB5 tok/s4K ctx
dense
F0
397B291.6 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs
