Benchmark Almanac

WebDev Arena

Claude Opus 5.5 1,820 leads since Sep 2026 Fading

Two anonymous models build web apps from a user's prompt, and the user votes for the better one. Votes yield Elo-style ratings.

Code Arena replaced it in November 2025, and the old board stays up as WebDev Legacy.

Made by Arena (formerly LMArena) Released Dec 2024 Web and app front ends Last in a launch post Oct 2026 , dropped for a newer version Last model added Oct 2026 Scores refreshed 31 Oct 2026
Scored
Ranked by votes People pick the better of two answers, as in an arena, and the votes are turned into ratings.
Answers
Open-ended Free-form text, documents, designs or media judged as a whole.
Leaderboard arena.ai
105
Models scored in Epoch AI data
2
Launch posts list it, from 1 lab
1
System card reports it, from 1 lab

Claude Opus 5.5 leads with 1,820

Best score per model in Epoch AI’s data. Where a model ran with several settings or harnesses, we keep its best.

105 models. Scroll sideways for the rest. Bars start at 1,000.

  1. 1,820 Claude Opus 5.5
  2. 1,800 GPT-6 Astra
  3. 1,786 Claude Sonnet 5.5
  4. 1,759 GPT-6.1 Sol
  5. 1,758 Claude Fable 5.1
  6. 1,689 GPT-6 Sol
  7. 1,687 Claude Opus 5
  8. 1,680 Gemini 4 Argon
  9. 1,674 Kimi K3
  10. 1,671 Qwen 3.8 Max
  11. 1,652 Muse Spark 1.3
  12. 1,638 Grok 4.7
  13. 1,628 Claude Fable 5
  14. 1,618 MiMo-V2.6 Pro
  15. 1,618 Grok 4.6
  16. 1,617 GPT-5.6 Sol
  17. 1,614 GLM-5.3
  18. 1,614 DeepSeek V4.1 Flash
  19. 1,607 GLM-5.3-Flash
  20. 1,593 Qwen 3.8 27B
  21. 1,593 GLM-5.2
  22. 1,587 Gemini 3.7 Flash
  23. 1,582 GPT-6 Luna
  24. 1,581 DeepSeek V4 Pro 0813
  25. 1,580 DeepSeek V4 Flash 0731
  26. 1,570 Step 5 Preview
  27. 1,568 Gemini 3.8 Flash
  28. 1,567 Claude Opus 4.7
  29. 1,559 Claude Opus 4.8
  30. 1,556 Claude Opus 4.6
  31. 1,555 Grok 4.5
  32. 1,542 Muse Spark 1.1
  33. 1,537 Gemini 3.6 Flash
  34. 1,537 Claude Sonnet 5
  35. 1,534 Muse Spark 1.2
  36. 1,521 GPT-5.6 Terra
  37. 1,521 Claude Sonnet 4.6
  38. 1,519 GPT-5.6 Luna
  39. 1,517 Qwen3.7-Max
  40. 1,513 Muse Spark
  41. 1,510 GPT-5.5
  42. 1,509 Kimi K2.6
  43. 1,508 GLM-5.1
  44. 1,506 Gemini 3.5 Flash
  45. 1,495 Claude Opus 4.5
  46. 1,487 MiniMax-M3
  47. 1,480 GPT-5.2
  48. 1,479 Qwen 3.6 Max (Preview)
  49. 1,475 MiMo-V2.5-Pro
  50. 1,473 Kimi K2.7 Code
  51. 1,464 DeepSeek-V4-Pro
  52. 1,463 GPT-5.4
  53. 1,461 Qwen 3.6 Plus
  54. 1,447 Gemini 3.1 Pro
  55. 1,447 Gemini 3.5 Flash-Lite
  56. 1,439 Gemini 3 Pro
  57. 1,437 MiMo-V2.5
  58. 1,436 Kimi K2.5
  59. 1,436 GLM-5
  60. 1,435 GLM-4.7
  61. 1,433 MiMo-V2-Pro
  62. 1,431 DeepSeek-V4-Flash
  63. 1,420 GPT-5
  64. 1,410 Inkling
  65. 1,407 Inkling-Small
  66. 1,399 Qwen 3.5 Plus (hosted 397B-A17B)
  67. 1,398 MiniMax-M2.7
  68. 1,397 GPT-5.4 Mini
  69. 1,392 GPT-5.1
  70. 1,391 Claude Sonnet 4.5
  71. 1,389 Claude Opus 4.1
  72. 1,387 MiniMax-M2.1
  73. 1,384 MiniMax-M2.5
  74. 1,383 Gemini 3 Flash
  75. 1,374 Grok 4.20
  76. 1,370 GPT-5.3 Codex
  77. 1,367 DeepSeek-V3.2-Exp
  78. 1,364 Gemma 4 31B IT
  79. 1,361 Gemma 4 26B A4B
  80. 1,357 Qwen3.5-27B
  81. 1,357 Grok 4.3 Beta
  82. 1,347 Laguna M.1
  83. 1,341 GLM-4.6
  84. 1,339 GPT-5.2 Codex
  85. 1,337 Kimi K2 Thinking
  86. 1,336 GPT-5.1-Codex
  87. 1,331 MiMo-V2-Flash
  88. 1,329 Claude Haiku 4.5
  89. 1,325 DeepSeek-V3.2
  90. 1,302 Laguna XS.2
  91. 1,298 MiniMax-M2
  92. 1,274 Qwen3-Coder-480B-A35B
  93. 1,265 Mistral Medium 3.5
  94. 1,254 Gemini 3.1 Flash-Lite
  95. 1,250 Qwen 3.6 35B-A3B
  96. 1,247 GPT-5.1-Codex-mini
  97. 1,238 Qwen 3.5 Flash (hosted 35B-A3B)
  98. 1,237 Trinity Large Thinking
  99. 1,230 Mistral Large 3
  100. 1,226 Gemini 2.5 Pro
  101. 1,223 Grok 4.1 Fast
  102. 1,211 Grok 4.1
  103. 1,167 Mercury 2
  104. 1,165 grok-code-fast-1
  105. 1,149 Grok 4 Fast

Bars start at 1,000.

  1. 1 Claude Opus 5.5 max 1,820
  2. 2 GPT-6 Astra max 1,800
  3. 3 Claude Sonnet 5.5 xhigh 1,786
  4. 4 GPT-6.1 Sol max 1,759
  5. 5 Claude Fable 5.1 max 1,758
  6. 6 GPT-6 Sol max 1,689
  7. 7 Claude Opus 5 max 1,687
  8. 8 Gemini 4 Argon high 1,680
  9. 9 Kimi K3 max 1,674
  10. 10 Qwen 3.8 Max 1,671
  11. 11 Muse Spark 1.3 max 1,652
  12. 12 Grok 4.7 xhigh 1,638
  13. 13 Claude Fable 5 high 1,628
  14. 14 MiMo-V2.6 Pro 1,618
  15. 15 Grok 4.6 high 1,618
  16. 16 GPT-5.6 Sol xhigh 1,617
  17. 17 GLM-5.3 max 1,614
  18. 18 DeepSeek V4.1 Flash max 1,614
  19. 19 GLM-5.3-Flash 1,607
  20. 20 Qwen 3.8 27B 1,593
  21. 21 GLM-5.2 max 1,593
  22. 22 Gemini 3.7 Flash high 1,587
  23. 23 GPT-6 Luna max 1,582
  24. 24 DeepSeek V4 Pro 0813 high 1,581
  25. 25 DeepSeek V4 Flash 0731 high 1,580
  26. 26 Step 5 Preview high 1,570
  27. 27 Gemini 3.8 Flash high 1,568
  28. 28 Claude Opus 4.7 1,567
  29. 29 Claude Opus 4.8 high 1,559
  30. 30 Claude Opus 4.6 unknown thinking 1,556
  31. 31 Grok 4.5 unknown thinking 1,555
  32. 32 Muse Spark 1.1 1,542
  33. 33 Gemini 3.6 Flash high 1,537
  34. 34 Claude Sonnet 5 high 1,537
  35. 35 Muse Spark 1.2 xhigh 1,534
  36. 36 GPT-5.6 Terra xhigh 1,521
  37. 37 Claude Sonnet 4.6 no thinking 1,521
  38. 38 GPT-5.6 Luna xhigh 1,519
  39. 39 Qwen3.7-Max 1,517
  40. 40 Muse Spark 1,513
  41. 41 GPT-5.5 xhigh 1,510
  42. 42 Kimi K2.6 1,509
  43. 43 GLM-5.1 1,508
  44. 44 Gemini 3.5 Flash 1,506
  45. 45 Claude Opus 4.5 32k thinking 1,495
  46. 46 MiniMax-M3 1,487
  47. 47 GPT-5.2 high 1,480
  48. 48 Qwen 3.6 Max (Preview) 1,479
  49. 49 MiMo-V2.5-Pro 1,475
  50. 50 Kimi K2.7 Code 1,473
  51. 51 DeepSeek-V4-Pro 1,464
  52. 52 GPT-5.4 high 1,463
  53. 53 Qwen 3.6 Plus 2026-04-02 1,461
  54. 54 Gemini 3.1 Pro 1,447
  55. 55 Gemini 3.5 Flash-Lite 1,447
  56. 56 Gemini 3 Pro 1,439
  57. 57 MiMo-V2.5 1,437
  58. 58 Kimi K2.5 1,436
  59. 59 GLM-5 1,436
  60. 60 GLM-4.7 1,435
  61. 61 MiMo-V2-Pro 1,433
  62. 62 DeepSeek-V4-Flash high 1,431
  63. 63 GPT-5 medium 1,420
  64. 64 Inkling 1,410
  65. 65 Inkling-Small 1,407
  66. 66 Qwen 3.5 Plus (hosted 397B-A17B) 1,399
  67. 67 MiniMax-M2.7 1,398
  68. 68 GPT-5.4 Mini high 1,397
  69. 69 GPT-5.1 medium 1,392
  70. 70 Claude Sonnet 4.5 32k thinking 1,391
  71. 71 Claude Opus 4.1 unknown thinking 1,389
  72. 72 MiniMax-M2.1 1,387
  73. 73 MiniMax-M2.5 1,384
  74. 74 Gemini 3 Flash 1,383
  75. 75 Grok 4.20 1,374
  76. 76 GPT-5.3 Codex 1,370
  77. 77 DeepSeek-V3.2-Exp thinking 1,367
  78. 78 Gemma 4 31B IT 1,364
  79. 79 Gemma 4 26B A4B 1,361
  80. 80 Qwen3.5-27B 1,357
  81. 81 Grok 4.3 Beta 1,357
  82. 82 Laguna M.1 1,347
  83. 83 GLM-4.6 1,341
  84. 84 GPT-5.2 Codex 1,339
  85. 85 Kimi K2 Thinking 1,337
  86. 86 GPT-5.1-Codex 1,336
  87. 87 MiMo-V2-Flash 1,331
  88. 88 Claude Haiku 4.5 unknown thinking 1,329
  89. 89 DeepSeek-V3.2 Thinking; Novita 1,325
  90. 90 Laguna XS.2 1,302
  91. 91 MiniMax-M2 1,298
  92. 92 Qwen3-Coder-480B-A35B 1,274
  93. 93 Mistral Medium 3.5 1,265
  94. 94 Gemini 3.1 Flash-Lite 1,254
  95. 95 Qwen 3.6 35B-A3B 1,250
  96. 96 GPT-5.1-Codex-mini 1,247
  97. 97 Qwen 3.5 Flash (hosted 35B-A3B) 1,238
  98. 98 Trinity Large Thinking 1,237
  99. 99 Mistral Large 3 1,230
  100. 100 Gemini 2.5 Pro Jun 2025 1,226
  101. 101 Grok 4.1 Fast 1,223
  102. 102 Grok 4.1 1,211
  103. 103 Mercury 2 1,167
  104. 104 grok-code-fast-1 1,165
  105. 105 Grok 4 Fast 1,149
#ModelScoreSettingDateSource
1 Claude Opus 5.5 1,820 max Sep 2026 Epoch AI
2 GPT-6 Astra 1,800 max Sep 2026 Epoch AI
3 Claude Sonnet 5.5 1,786 xhigh Sep 2026 Epoch AI
4 GPT-6.1 Sol 1,759 max Sep 2026 Epoch AI
5 Claude Fable 5.1 1,758 max Sep 2026 Epoch AI
6 GPT-6 Sol 1,689 max Sep 2026 Epoch AI
7 Claude Opus 5 1,687 max Jul 2026 Epoch AI
8 Gemini 4 Argon 1,680 high Sep 2026 Epoch AI
9 Kimi K3 1,674 max Jul 2026 Epoch AI
10 Qwen 3.8 Max 1,671 Aug 2026 Epoch AI
11 Muse Spark 1.3 1,652 max Sep 2026 Epoch AI
12 Grok 4.7 1,638 xhigh Sep 2026 Epoch AI
13 Claude Fable 5 1,628 high Jun 2026 Epoch AI
14 MiMo-V2.6 Pro 1,618 Sep 2026 Epoch AI
15 Grok 4.6 1,618 high Aug 2026 Epoch AI
16 GPT-5.6 Sol 1,617 xhigh Jul 2026 Epoch AI
17 GLM-5.3 1,614 max Aug 2026 Epoch AI
18 DeepSeek V4.1 Flash 1,614 max Sep 2026 Epoch AI
19 GLM-5.3-Flash 1,607 Aug 2026 Epoch AI
20 Qwen 3.8 27B 1,593 Aug 2026 Epoch AI
21 GLM-5.2 1,593 max Jun 2026 Epoch AI
22 Gemini 3.7 Flash 1,587 high Aug 2026 Epoch AI
23 GPT-6 Luna 1,582 max Sep 2026 Epoch AI
24 DeepSeek V4 Pro 0813 1,581 high Aug 2026 Epoch AI
25 DeepSeek V4 Flash 0731 1,580 high Jul 2026 Epoch AI
26 Step 5 Preview 1,570 high Sep 2026 Epoch AI
27 Gemini 3.8 Flash 1,568 high Sep 2026 Epoch AI
28 Claude Opus 4.7 1,567 Apr 2026 Epoch AI
29 Claude Opus 4.8 1,559 high May 2026 Epoch AI
30 Claude Opus 4.6 1,556 unknown thinking Feb 2026 Epoch AI
31 Grok 4.5 1,555 unknown thinking Jul 2026 Epoch AI
32 Muse Spark 1.1 1,542 Jul 2026 Epoch AI
33 Gemini 3.6 Flash 1,537 high Jul 2026 Epoch AI
34 Claude Sonnet 5 1,537 high Jun 2026 Epoch AI
35 Muse Spark 1.2 1,534 xhigh Aug 2026 Epoch AI
36 GPT-5.6 Terra 1,521 xhigh Jul 2026 Epoch AI
37 Claude Sonnet 4.6 1,521 no thinking Feb 2026 Epoch AI
38 GPT-5.6 Luna 1,519 xhigh Jul 2026 Epoch AI
39 Qwen3.7-Max 1,517 May 2026 Epoch AI
40 Muse Spark 1,513 Apr 2026 Epoch AI
41 GPT-5.5 1,510 xhigh Apr 2026 Epoch AI
42 Kimi K2.6 1,509 Apr 2026 Epoch AI
43 GLM-5.1 1,508 Apr 2026 Epoch AI
44 Gemini 3.5 Flash 1,506 May 2026 Epoch AI
45 Claude Opus 4.5 1,495 32k thinking Nov 2025 Epoch AI
46 MiniMax-M3 1,487 Jun 2026 Epoch AI
47 GPT-5.2 1,480 high Dec 2025 Epoch AI
48 Qwen 3.6 Max (Preview) 1,479 Apr 2026 Epoch AI
49 MiMo-V2.5-Pro 1,475 Apr 2026 Epoch AI
50 Kimi K2.7 Code 1,473 Jun 2026 Epoch AI
51 DeepSeek-V4-Pro 1,464 Apr 2026 Epoch AI
52 GPT-5.4 1,463 high Mar 2026 Epoch AI
53 Qwen 3.6 Plus 1,461 2026-04-02 Mar 2026 Epoch AI
54 Gemini 3.1 Pro 1,447 Feb 2026 Epoch AI
55 Gemini 3.5 Flash-Lite 1,447 Jul 2026 Epoch AI
56 Gemini 3 Pro 1,439 Nov 2025 Epoch AI
57 MiMo-V2.5 1,437 Apr 2026 Epoch AI
58 Kimi K2.5 1,436 Jan 2026 Epoch AI
59 GLM-5 1,436 Feb 2026 Epoch AI
60 GLM-4.7 1,435 Dec 2025 Epoch AI
61 MiMo-V2-Pro 1,433 Mar 2026 Epoch AI
62 DeepSeek-V4-Flash 1,431 high Apr 2026 Epoch AI
63 GPT-5 1,420 medium Aug 2025 Epoch AI
64 Inkling 1,410 Jul 2026 Epoch AI
65 Inkling-Small 1,407 Jul 2026 Epoch AI
66 Qwen 3.5 Plus (hosted 397B-A17B) 1,399 Feb 2026 Epoch AI
67 MiniMax-M2.7 1,398 Mar 2026 Epoch AI
68 GPT-5.4 Mini 1,397 high Mar 2026 Epoch AI
69 GPT-5.1 1,392 medium Nov 2025 Epoch AI
70 Claude Sonnet 4.5 1,391 32k thinking Sep 2025 Epoch AI
71 Claude Opus 4.1 1,389 unknown thinking Aug 2025 Epoch AI
72 MiniMax-M2.1 1,387 Dec 2025 Epoch AI
73 MiniMax-M2.5 1,384 Feb 2026 Epoch AI
74 Gemini 3 Flash 1,383 Dec 2025 Epoch AI
75 Grok 4.20 1,374 Feb 2026 Epoch AI
76 GPT-5.3 Codex 1,370 Feb 2026 Epoch AI
77 DeepSeek-V3.2-Exp 1,367 thinking Sep 2025 Epoch AI
78 Gemma 4 31B IT 1,364 Apr 2026 Epoch AI
79 Gemma 4 26B A4B 1,361 Apr 2026 Epoch AI
80 Qwen3.5-27B 1,357 Feb 2026 Epoch AI
81 Grok 4.3 Beta 1,357 Apr 2026 Epoch AI
82 Laguna M.1 1,347 Apr 2026 Epoch AI
83 GLM-4.6 1,341 Sep 2025 Epoch AI
84 GPT-5.2 Codex 1,339 Dec 2025 Epoch AI
85 Kimi K2 Thinking 1,337 Nov 2025 Epoch AI
86 GPT-5.1-Codex 1,336 Nov 2025 Epoch AI
87 MiMo-V2-Flash 1,331 Dec 2025 Epoch AI
88 Claude Haiku 4.5 1,329 unknown thinking Oct 2025 Epoch AI
89 DeepSeek-V3.2 1,325 Thinking; Novita Dec 2025 Epoch AI
90 Laguna XS.2 1,302 Apr 2026 Epoch AI
91 MiniMax-M2 1,298 Oct 2025 Epoch AI
92 Qwen3-Coder-480B-A35B 1,274 Jul 2025 Epoch AI
93 Mistral Medium 3.5 1,265 Apr 2026 Epoch AI
94 Gemini 3.1 Flash-Lite 1,254 Mar 2026 Epoch AI
95 Qwen 3.6 35B-A3B 1,250 Apr 2026 Epoch AI
96 GPT-5.1-Codex-mini 1,247 Nov 2025 Epoch AI
97 Qwen 3.5 Flash (hosted 35B-A3B) 1,238 Feb 2026 Epoch AI
98 Trinity Large Thinking 1,237 Apr 2026 Epoch AI
99 Mistral Large 3 1,230 Dec 2025 Epoch AI
100 Gemini 2.5 Pro Jun 2025 1,226 Jun 2025 Epoch AI
101 Grok 4.1 Fast 1,223 Nov 2025 Epoch AI
102 Grok 4.1 1,211 Nov 2025 Epoch AI
103 Mercury 2 1,167 Feb 2026 Epoch AI
104 grok-code-fast-1 1,165 Aug 2025 Epoch AI
105 Grok 4 Fast 1,149 Sep 2025 Epoch AI

Labs report scores for 2 models in their cards. We plot every source in the charts below.

14 models have held number one since its release in Dec 2024

Lab claims count from the day the card came out, independent runs from the model’s release.

Google DeepMind Qwen Anthropic OpenAI Moonshot AI
Epoch AI runs 14 leaders
Claude Opus 5.5 1,820

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Run or collected by Epoch AI Reported by the lab in its model card
1,000 1,200 1,400 1,600 1,800 2,000 2025 2026 2027 1,000 1,200 1,400 1,600 1,800 2,000 2025 2026 2027

Labs headline it less often now, last in Oct 2026

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Google DeepMind
DateLaunch postScoreAs the post writes it
28 Oct 2026 Gemini 4 Carbon 1768 WebDev Arena (#4, 1768, text)
6 May 2025 Gemini 2.5 Pro 1420 WebDev Arena Leaderboard 05-06-25, Arena Score (#1)

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
7 Jul 2025 Gemini 2.5 Pro technical report LMArena WebDev Arena, text says an increase of over 500 Elo over Gemini 1.5 Pro, no absolute score (Section 2.6) Capabilities

Versions

Replaced by Code Arena

More in Coding

All 243 in Coding
  • SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Opus 5.5 93.9%
  • LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests. DeepSeek-V4-Pro 93.5%
  • DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
  • SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases. Seed2.1 Pro 59.8%
  • Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance. DeepSeek V4.1 Flash 3,471