Local-first ranking

The Local AI Leaderboard

Artificial Analysis ranks raw frontier intelligence - and almost all of it is proprietary, cloud-only, and metered by the token. We flip it: this board ranks the best models you can actually self-host, defaulting to open weights and surfacing the VRAM and params you need to run them. Same Intelligence / Coding / Math / Agentic indexes - reframed for $0/mo local instead of pay-per-call cloud.

Best you can run locally

Kimi K3 leads the open field at 50.2 Intelligence - ~1612 GB VRAM to run. Buy the GPU once, then $0/mo forever. No tokens, no rate limits, your data stays home.

Frontier, but cloud-only

Claude Fable 5.1 (max with fallback) tops the proprietary board at 56.8 - but you can't download it. It's metered per token, can change or vanish overnight, and every prompt leaves your machine. Use the toggle below to compare it against what you can self-host.

open / runs locally · $0/mo proprietary - not local
1
Kimi K3
2779.9B~1612 GB
50.2
2
GLM-5.3
321.3B~186 GB
48.6
3
Qwen3.8 2.4T A95B
2446.2B~1419 GB
46.7
4
GLM-5.3-Flash
321.3B~186 GB
46.2
5
Qwen3.8-Flash-Next
180B~104 GB
45.6
6
GLM-5.2
753.4B~437 GB
42.5
7
DeepSeek V4 Pro 0813
1650.5B~957 GB
42.1
8
Qwen3.8 27B
27.8B~16 GB
41.6
9
DeepSeek V4 Flash 0731
304.2B~176 GB
40.8
10
GLM-5.1
744B
40.2
11
MiniMax-M2.7
228.7B~133 GB
38.1
12
Motif 3
314.8B~183 GB
38.0
13
DeepSeek V4 Pro
1600B
35.8
14
MiniMax-M3
440.3B~255 GB
35.7
15
Kimi K2.6
1000B
34.6
16
Kimi K2.7 Code
1058.6B~614 GB
33.9
17
Hy3
298.8B~173 GB
33.2
18
Nex-N2-Pro
396.8B~230 GB
32.7
19
MiMo-V2.5-Pro
1023.2B~593 GB
32.6
20
Inkling Small
156B~90 GB
32.3
21
Inkling
552.8B~321 GB
32.2
22
Nemotron 3 Ultra
253.4B~147 GB
29.6
23
MiMo-V2.5
310.8B~180 GB
29.5
24
Qwen3.6 27B
28B~16 GB
29.0
25
A.X-K2
21.2B~12 GB
26.8
26
Qwen3.6 35B A3B
36B~21 GB
26.2
27
Qwen3.5 397B A17B
403.4B~234 GB
26.1
28
G9v3-39A5B
39B~23 GB
25.7
29
NVIDIA Nemotron 3 Super
67.2B~39 GB
25.4
30
Qwen3.5 122B A10B
125.1B~73 GB
24.8
31
Muse Glimmer
18.1B~10 GB
24.4
32
Gemma 4 31B
32.7B~19 GB
22.2
33
DeepSeek V4 Flash
284B
22.1
34
Gemma 4 26B A4B
26.5B~15 GB
19.2
35
Nemotron 3 Super
67.2B~39 GB
18.6
36
Qwen3.5 35B A3B
36B~21 GB
17.0
37
Nemotron 3.5 Lightning
17.8B~10 GB
16.4
38
Ling 3.0 Tiny
7.9B~5 GB
15.7
39
Gemma 4 12B
12B~7 GB
15.6
40
gpt-oss-120b
120.4B~70 GB
15.6
41
Qwen3.5 9B
9.7B~6 GB
14.8
42
Qwen3-Coder-Next
79.7B~46 GB
14.3
43
EXAONE 4.5 33B
34.4B~20 GB
14.0
44
Qwen3.5 4B
4.7B~3 GB
13.9
45
North Mini Code
30.5B~18 GB
13.5
46
HyperNova 60B 2605
58.7B~34 GB
11.7
47
Nemotron Cascade 2 30B A3B
31.6B~18 GB
11.5
48
Qwen3-Next-80B-A3B-Instruct
81.3B~47 GB
10.7
49
gpt-oss-20b
21.5B~12 GB
9.1
50
MiniCPM5-1B
1.1B~1 GB
6.2
51
Llama 3.3 70B
70B~40 GB
3.7
52
NVIDIA-Nemotron-Nano-9B-v2
8.9B~5 GB
3.2
53
granite-4.1-8b
8.8B~5 GB
1.1
54
gemma-3-270m
0.3B
1.0
55
granite-4.1-3b
3.4B~2 GB
1.0
56
LFM2.5-1.2B-Instruct
1.2B~1 GB
1.0
57
Phi-4 Mini
3B~2.5 GB
1.0
58
Phi-4
14B~8.5 GB
1.0
58 open models you can run locally · sorted by Intelligence IndexData: Artificial Analysis · updated 2026-09-05

Found a model that fits your GPU? Each one has a full page with benchmarks, VRAM math, and recommended hardware.

Browse all local models