What is the best open LLM on LMArena WebDev?

Kimi K3 is the top open model on LMArena WebDev, scoring 1657.9. Among all models tested — including proprietary ones — it ranks #7. The top model overall is GPT 6 Astra Max (OpenAI) at 1793.2.

What's the best LMArena WebDev model you can run on a 24 GB GPU?

Qwen3.8 27B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (about 15 GB), scoring 1591.2 on LMArena WebDev.

What's the best LMArena WebDev model you can run on a 12 GB GPU?

Granite 4.1 8B is the highest-scoring open model that fits in 12 GB at 4-bit quantization (about 5 GB), scoring 1191.6 on LMArena WebDev.

Can open models match proprietary models on LMArena WebDev?

Not quite on LMArena WebDev: the strongest proprietary model (GPT 6 Astra Max) scores 1793.2, ahead of the best open model (Kimi K3) at 1657.9 — but you can run the open one yourself.

Coding

LMArena WebDev Leaderboard

Name: LMArena WebDev — open LLM scores
Creator: lmarena

Code Arena (formerly WebDev Arena) is Arena's head-to-head coding leaderboard: two models each build a real web app or UI from the same prompt, and people vote blind on which result is better. The rating shown is a relative Bradley-Terry score, not a percentage — compare it only to other models on this board.

Source: lmarena48 open models ranked+75 proprietaryData through Sep 2026

Open models All models

Open models ranked on LMArena WebDev

# shows rank among open models / rank overall (including proprietary).

#	Model	Score
1 / 7	Kimi K3 · 2779.9B	1657.9
2 / 9	Qwen3.8 Flash Next · 180.0B	1635.6
3 / 11	Hy4 Preview · 780.0B	1627.4
4 / 14	GLM 5.3 · 753.3B	1619.7
5 / 17	DeepSeek V4.1 Flash · 763.2B	1616.1
6 / 18	GLM 5.3 Flash · 321.3B	1612.2
7 / 19	GLM 5.2 · 753.3B	1598.2
8 / 21	Qwen3.8 27B · 27.8B	1591.2
9 / 23	DeepSeek V4 Pro 0813 · 1650.5B	1580.2
10 / 24	DeepSeek V4 Flash · 290.9B	1579.9
11 / 42	Hy3 · 298.8B	1508.7
12 / 43	Kimi K2.6 · 1026.9B	1508.6
13 / 44	GLM 5.1 · 753.9B	1508.1
14 / 49	MiniMax M3 · 427.0B	1485.0
15 / 51	MiMo V2.5 Pro · 1023.2B	1476.8
16 / 52	Kimi K2.7 Code · 1026.9B	1472.3
17 / 54	DeepSeek V4 Pro · 1598.8B	1463.5
18 / 63	MiMo V2.5 · 310.8B	1437.4
19 / 64	Kimi K2.5 · 1026.9B	1436.3
20 / 65	GLM 5 · 753.9B	1435.8
21 / 66	GLM 4.7 · 358.3B	1434.9
22 / 70	Inkling · 952.4B	1411.1
23 / 72	Inkling Small · 266.0B	1406.7
24 / 74	Qwen3.5 397B A17B · 403.4B	1398.8
25 / 75	MiniMax M2.7 · 228.7B	1397.5
26 / 81	MiniMax M2.1 · 228.7B	1387.4
27 / 83	MiniMax M2.5 · 228.7B	1384.0
28 / 87	Gemma 4 31B IT · 31.3B	1364.2
29 / 88	Gemma 4 26B A4B IT · 25.8B	1361.9
30 / 89	DeepSeek V3.2 · 685.4B	1360.8
31 / 90	Muse Glimmer 30B · 29.8B	1359.8
32 / 91	Qwen3.5 122B A10B · 125.1B	1358.2
33 / 92	Qwen3.5 27B · 27.8B	1357.8
34 / 93	Hy3 Preview · 298.8B	1357.0
35 / 95	Laguna M.1 · 225.8B	1347.2
36 / 97	GLM 4.6 · 356.8B	1340.8
37 / 100	MiMo v2 Flash · 309.8B	1330.7
38 / 102	Kimi K2 Thinking · 1026.4B	1323.0
39 / 103	Laguna XS.2 · 33.4B	1302.3
40 / 104	MiniMax M2 · 228.7B	1298.0
41 / 105	Qwen3 Coder 480B A35B Instruct · 480.2B	1273.9
42 / 106	DeepSeek V3.2 Exp · 685.4B	1271.8
43 / 107	Mistral Medium 3.5 128B · 127.7B	1264.6
44 / 110	Qwen3.5 35B A3B · 36.0B	1250.3
45 / 114	Trinity Large Thinking · 398.6B	1236.8
46 / 115	Mistral Large 3 675B Instruct 2512 · 675B	1228.8
47 / 118	Devstral 2 123B Instruct 2512 · 125.0B	1194.2
48 / 119	Granite 4.1 8B · 8.8B	1191.6

Score vs model size

Which models give the most quality for their size — the ones worth running locally.

Each dot is a model. Up = higher score, left = smaller (easier to run locally). The dashed line marks the efficiency frontier — the best score you can get at each size or smaller.

LMArena WebDev: frequently asked questions

What is the best open LLM on LMArena WebDev?: Kimi K3 is the top open model on LMArena WebDev, scoring 1657.9. Among all models tested — including proprietary ones — it ranks #7. The top model overall is GPT 6 Astra Max (OpenAI) at 1793.2.
What's the best LMArena WebDev model you can run on a 24 GB GPU?: Qwen3.8 27B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (about 15 GB), scoring 1591.2 on LMArena WebDev.
What's the best LMArena WebDev model you can run on a 12 GB GPU?: Granite 4.1 8B is the highest-scoring open model that fits in 12 GB at 4-bit quantization (about 5 GB), scoring 1191.6 on LMArena WebDev.
Can open models match proprietary models on LMArena WebDev?: Not quite on LMArena WebDev: the strongest proprietary model (GPT 6 Astra Max) scores 1793.2, ahead of the best open model (Kimi K3) at 1657.9 — but you can run the open one yourself.

Scores aggregated from lmarena. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.