Benchmark Almanac

Code Arena

Claude Opus 5.5 1,810 leads since Sep 2026 In use

Models act as coding agents that build web apps from user prompts, and users vote between two results. Votes yield Elo-style ratings.

It started a fresh leaderboard and does not reuse WebDev Arena votes. It splits results into Fullstack and Frontend categories, with HTML, React and domain filters. On 2026-10-27 it listed 153 models and 952,039 votes. Image-to-WebDev is a separate board.

Made by Arena (formerly LMArena) Released Nov 2025 Web and app front ends Last in a launch post Aug 2026 Last model added Oct 2026 Scores refreshed 2 Nov 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
129
Entries on its own leaderboard, 124 models we track
1
Launch post lists it, from 1 lab
3
System cards report it, from 3 labs

Claude Opus 5.5 leads with 1,810

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

129 entries. Scroll sideways for the rest. Bars start at 1,000.

  1. 1,810 Claude Opus 5.5
  2. 1,773 GPT-6 Astra
  3. 1,770 Claude Sonnet 5.5
  4. 1,768 Gemini 4 Carbon
  5. 1,751 GPT-6.1 Sol
  6. 1,746 Claude Fable 5.1
  7. 1,725 Muse Wonder Full
  8. 1,695 Kimi K3.1 Pro
  9. 1,691 Claude Opus 5
  10. 1,679 Gemini 4 Argon
  11. 1,678 GPT-6 Sol
  12. 1,676 Qwen3.8 Max (0902)
  13. 1,673 GPT-6.1 Terra
  14. 1,672 Qwen 3.8 Max
  15. 1,665 DeepSeek V4.1 Flash
  16. 1,656 Kimi K3
  17. 1,656 Muse Spark 1.3
  18. 1,652 Muse Wonder
  19. 1,637 Grok 4.7
  20. 1,637 MiMo-V2.6 Flash
  21. 1,631 Tencent Hy4 preview
  22. 1,628 Qwen3.8-Flash-Next
  23. 1,625 MiMo-V2.6 Pro
  24. 1,624 Claude Fable 5
  25. 1,623 GLM-5.3
  26. 1,616 Grok 4.6
  27. 1,616 GPT-5.6 Sol
  28. 1,607 GLM-5.3-Flash
  29. 1,606 GLM-5.2
  30. 1,590 Qwen 3.8 27B
  31. 1,590 Gemini 3.7 Flash
  32. 1,589 Claude Haiku 5.5
  33. 1,583 Gemini 3.8 Flash
  34. 1,582 MiniMax-M3.1-Flash
  35. 1,582 DeepSeek-V4-Pro
  36. 1,582 DeepSeek-V4-Flash
  37. 1,579 GPT-6 Luna
  38. 1,571 Step 5 Preview
  39. 1,558 Claude Opus 4.7
  40. 1,556 Claude Opus 4.8
  41. 1,553 Grok 4.5
  42. 1,546 Claude Opus 4.6
  43. 1,544 Mistral Large 4
  44. 1,542 Muse Spark 1.1
  45. 1,540 Claude Sonnet 5
  46. 1,538 Gemini 3.6 Flash
  47. 1,534 Muse Spark 1.2
  48. 1,526 Muse Spark Open
  49. 1,523 Claude Sonnet 4.6
  50. 1,520 GPT-5.6 Terra
  51. 1,519 Seed2.1 Pro
  52. 1,519 GPT-5.6 Luna
  53. 1,515 Qwen3.7-Max
  54. 1,512 GPT-5.5
  55. 1,509 Kimi K2.6
  56. 1,508 GLM-5.1
  57. 1,505 Tencent Hy3
  58. 1,499 Gemini 3.5 Flash
  59. 1,494 Claude Opus 4.5
  60. 1,483 MiniMax-M3
  61. 1,482 Spark-X2.6
  62. 1,482 Qwen 3.6 Max (Preview)
  63. 1,480 MiMo-V2.5-Pro
  64. 1,474 Kimi K2.7 Code
  65. 1,463 GPT-5.4
  66. 1,462 Qwen 3.6 Plus
  67. 1,447 Gemini 3.1 Pro
  68. 1,440 Gemini 3 Flash
  69. 1,440 Gemini 3 Pro
  70. 1,438 MiMo-V2.5
  71. 1,436 Kimi K2.5
  72. 1,436 Beam 1.0
  73. 1,436 GLM-5
  74. 1,435 GLM-4.7
  75. 1,434 MiMo-V2-Pro
  76. 1,434 Gemini 3.5 Flash-Lite
  77. 1,417 GPT-5
  78. 1,416 GPT-5.2
  79. 1,415 Inkling
  80. 1,409 Inkling-Small
  81. 1,407 GPT-5.3 Codex
  82. 1,401 GLM-5V-Turbo
  83. 1,400 Qwen3.5 397B-A17B
  84. 1,399 Nemotron 3.5 Super VL
  85. 1,398 MiniMax-M2.7
  86. 1,397 GPT-5.4 Mini
  87. 1,396 GPT-5.1
  88. 1,394 Claude Sonnet 4.5
  89. 1,390 Claude Opus 4.1
  90. 1,386 MiniMax-M2.5
  91. 1,384 MiniMax-M2.1
  92. 1,375 Grok 4.20 Beta (Reasoning, 0309)
  93. 1,370 Solar Pro 4
  94. 1,369 Gemma 4 31B IT
  95. 1,363 DeepSeek-V3.2
  96. 1,360 Qwen3.5-122B-A10B
  97. 1,359 Gemma 4 26B A4B
  98. 1,358 Qwen3.5-27B
  99. 1,358 Grok 4.3
  100. 1,357 Tencent Hy3 preview
  101. 1,355 Muse Glimmer
  102. 1,349 Laguna M.1
  103. 1,341 GLM-4.6
  104. 1,339 GPT-5.2 Codex
  105. 1,337 GPT-5.1-Codex
  106. 1,331 MiMo-V2-Flash
  107. 1,330 Claude Haiku 4.5
  108. 1,322 Kimi K2 Thinking Turbo
  109. 1,302 Laguna XS.2
  110. 1,297 MiniMax-M2
  111. 1,275 Qwen3-Coder-480B-A35B
  112. 1,271 DeepSeek-V3.2-Exp
  113. 1,264 Mistral Medium 3.5
  114. 1,256 Gemini 3.1 Flash-Lite
  115. 1,256 KAT-Coder-Pro V1
  116. 1,254 Qwen3.5-35B-A3B
  117. 1,244 Qwen3.5-Flash
  118. 1,244 GPT-5.1-Codex-mini
  119. 1,242 Grok 4.1 Fast
  120. 1,238 Trinity Large Thinking
  121. 1,233 Mistral Large 3
  122. 1,226 Gemini 2.5 Pro
  123. 1,214 Grok 4.1
  124. 1,197 Devstral 2
  125. 1,192 Granite 4.1 8B
  126. 1,170 Mercury 2
  127. 1,166 grok-code-fast-1
  128. 1,159 Grok 4 Fast
  129. 1,071 Devstral Medium

Bars start at 1,000.

  1. 1 Claude Opus 5.5 max 1,810
  2. 2 GPT-6 Astra max 1,773
  3. 3 Claude Sonnet 5.5 xhigh 1,770
  4. 4 Gemini 4 Carbon high 1,768
  5. 5 GPT-6.1 Sol max 1,751
  6. 6 Claude Fable 5.1 max 1,746
  7. 7 Muse Wonder Full max 1,725
  8. 8 Kimi K3.1 Pro max 1,695
  9. 9 Claude Opus 5 max 1,691
  10. 10 Gemini 4 Argon high 1,679
  11. 11 GPT-6 Sol max 1,678
  12. 12 Qwen3.8 Max (0902) 1,676
  13. 13 GPT-6.1 Terra max 1,673
  14. 14 Qwen 3.8 Max 1,672
  15. 15 DeepSeek V4.1 Flash max 1,665
  16. 16 Kimi K3 max 1,656
  17. 17 Muse Spark 1.3 max 1,656
  18. 18 Muse Wonder max 1,652
  19. 19 Grok 4.7 xhigh 1,637
  20. 20 MiMo-V2.6 Flash 1,637
  21. 21 Tencent Hy4 preview 1,631
  22. 22 Qwen3.8-Flash-Next 1,628
  23. 23 MiMo-V2.6 Pro 1,625
  24. 24 Claude Fable 5 high 1,624
  25. 25 GLM-5.3 max 1,623
  26. 26 Grok 4.6 high 1,616
  27. 27 GPT-5.6 Sol 1,616
  28. 28 GLM-5.3-Flash 1,607
  29. 29 GLM-5.2 max 1,606
  30. 30 Qwen 3.8 27B 1,590
  31. 31 Gemini 3.7 Flash high 1,590
  32. 32 Claude Haiku 5.5 high 1,589
  33. 33 Gemini 3.8 Flash high 1,583
  34. 34 MiniMax-M3.1-Flash 1,582
  35. 35 DeepSeek-V4-Pro high 1,582
  36. 36 DeepSeek-V4-Flash high 1,582
  37. 37 GPT-6 Luna max 1,579
  38. 38 Step 5 Preview high 1,571
  39. 39 Claude Opus 4.7 high 1,558
  40. 40 Claude Opus 4.8 high 1,556
  41. 41 Grok 4.5 1,553
  42. 42 Claude Opus 4.6 high 1,546
  43. 43 Mistral Large 4 1,544
  44. 44 Muse Spark 1.1 1,542
  45. 45 Claude Sonnet 5 high 1,540
  46. 46 Gemini 3.6 Flash high 1,538
  47. 47 Muse Spark 1.2 xhigh 1,534
  48. 48 Muse Spark Open max 1,526
  49. 49 Claude Sonnet 4.6 1,523
  50. 50 GPT-5.6 Terra 1,520
  51. 51 Seed2.1 Pro 1,519
  52. 52 GPT-5.6 Luna 1,519
  53. 53 Qwen3.7-Max 1,515
  54. 54 GPT-5.5 1,512
  55. 55 Kimi K2.6 1,509
  56. 56 GLM-5.1 1,508
  57. 57 Tencent Hy3 1,505
  58. 58 Gemini 3.5 Flash high 1,499
  59. 59 Claude Opus 4.5 high, 32k 1,494
  60. 60 MiniMax-M3 1,483
  61. 61 Spark-X2.6 1,482
  62. 62 Qwen 3.6 Max (Preview) 1,482
  63. 63 MiMo-V2.5-Pro 1,480
  64. 64 Kimi K2.7 Code 1,474
  65. 65 GPT-5.4 1,463
  66. 66 Qwen 3.6 Plus 1,462
  67. 67 Gemini 3.1 Pro 1,447
  68. 68 Gemini 3 Flash 1,440
  69. 69 Gemini 3 Pro 1,440
  70. 70 MiMo-V2.5 1,438
  71. 71 Kimi K2.5 thinking 1,436
  72. 72 Beam 1.0 1,436
  73. 73 GLM-5 1,436
  74. 74 GLM-4.7 1,435
  75. 75 MiMo-V2-Pro 1,434
  76. 76 Gemini 3.5 Flash-Lite 1,434
  77. 77 GPT-5 medium 1,417
  78. 78 GPT-5.2 1,416
  79. 79 Inkling 1,415
  80. 80 Inkling-Small 1,409
  81. 81 GPT-5.3 Codex 1,407
  82. 82 GLM-5V-Turbo 1,401
  83. 83 Qwen3.5 397B-A17B 1,400
  84. 84 Nemotron 3.5 Super VL nvfp4 1,399
  85. 85 MiniMax-M2.7 1,398
  86. 86 GPT-5.4 Mini high 1,397
  87. 87 GPT-5.1 medium 1,396
  88. 88 Claude Sonnet 4.5 high, 32k 1,394
  89. 89 Claude Opus 4.1 1,390
  90. 90 MiniMax-M2.5 1,386
  91. 91 MiniMax-M2.1 1,384
  92. 92 Grok 4.20 Beta (Reasoning, 0309) 1,375
  93. 93 Solar Pro 4 1,370
  94. 94 Gemma 4 31B IT 1,369
  95. 95 DeepSeek-V3.2 thinking 1,363
  96. 96 Qwen3.5-122B-A10B 1,360
  97. 97 Gemma 4 26B A4B 1,359
  98. 98 Qwen3.5-27B 1,358
  99. 99 Grok 4.3 1,358
  100. 100 Tencent Hy3 preview 1,357
  101. 101 Muse Glimmer 1,355
  102. 102 Laguna M.1 1,349
  103. 103 GLM-4.6 1,341
  104. 104 GPT-5.2 Codex 1,339
  105. 105 GPT-5.1-Codex 1,337
  106. 106 MiMo-V2-Flash non-thinking 1,331
  107. 107 Claude Haiku 4.5 1,330
  108. 108 Kimi K2 Thinking Turbo 1,322
  109. 109 Laguna XS.2 1,302
  110. 110 MiniMax-M2 1,297
  111. 111 Qwen3-Coder-480B-A35B 1,275
  112. 112 DeepSeek-V3.2-Exp 1,271
  113. 113 Mistral Medium 3.5 1,264
  114. 114 Gemini 3.1 Flash-Lite 1,256
  115. 115 KAT-Coder-Pro V1 1,256
  116. 116 Qwen3.5-35B-A3B 1,254
  117. 117 Qwen3.5-Flash 1,244
  118. 118 GPT-5.1-Codex-mini 1,244
  119. 119 Grok 4.1 Fast reasoning 1,242
  120. 120 Trinity Large Thinking 1,238
  121. 121 Mistral Large 3 1,233
  122. 122 Gemini 2.5 Pro Jun 2025 1,226
  123. 123 Grok 4.1 thinking 1,214
  124. 124 Devstral 2 1,197
  125. 125 Granite 4.1 8B 1,192
  126. 126 Mercury 2 1,170
  127. 127 grok-code-fast-1 1,166
  128. 128 Grok 4 Fast reasoning 1,159
  129. 129 Devstral Medium 1,071
#EntryScoreSettingDateSource
1 Claude Opus 5.5 1,810 max Sep 2026 Official leaderboard
2 GPT-6 Astra 1,773 max Sep 2026 Official leaderboard
3 Claude Sonnet 5.5 1,770 xhigh Sep 2026 Official leaderboard
4 Gemini 4 Carbon 1,768 high Oct 2026 Official leaderboard
5 GPT-6.1 Sol 1,751 max Sep 2026 Official leaderboard
6 Claude Fable 5.1 1,746 max Sep 2026 Official leaderboard
7 Muse Wonder Full 1,725 max Oct 2026 Official leaderboard
8 Kimi K3.1 Pro 1,695 max Oct 2026 Official leaderboard
9 Claude Opus 5 1,691 max Jul 2026 Official leaderboard
10 Gemini 4 Argon 1,679 high Sep 2026 Official leaderboard
11 GPT-6 Sol 1,678 max Sep 2026 Official leaderboard
12 Qwen3.8 Max (0902) 1,676 Sep 2026 Official leaderboard
13 GPT-6.1 Terra 1,673 max Oct 2026 Official leaderboard
14 Qwen 3.8 Max 1,672 Aug 2026 Official leaderboard
15 DeepSeek V4.1 Flash 1,665 max Sep 2026 Official leaderboard
16 Kimi K3 1,656 max Jul 2026 Official leaderboard
17 Muse Spark 1.3 1,656 max Sep 2026 Official leaderboard
18 Muse Wonder 1,652 max Oct 2026 Official leaderboard
19 Grok 4.7 1,637 xhigh Sep 2026 Official leaderboard
20 MiMo-V2.6 Flash 1,637 Sep 2026 Official leaderboard
21 Tencent Hy4 preview 1,631 Aug 2026 Official leaderboard
22 Qwen3.8-Flash-Next 1,628 Oct 2026 Official leaderboard
23 MiMo-V2.6 Pro 1,625 Sep 2026 Official leaderboard
24 Claude Fable 5 1,624 high Jun 2026 Official leaderboard
25 GLM-5.3 1,623 max Aug 2026 Official leaderboard
26 Grok 4.6 1,616 high Aug 2026 Official leaderboard
27 GPT-5.6 Sol 1,616 xHigh, Codex Jul 2026 Official leaderboard
28 GLM-5.3-Flash 1,607 Aug 2026 Official leaderboard
29 GLM-5.2 1,606 max Jun 2026 Official leaderboard
30 Qwen 3.8 27B 1,590 Aug 2026 Official leaderboard
31 Gemini 3.7 Flash 1,590 high Aug 2026 Official leaderboard
32 Claude Haiku 5.5 1,589 high Oct 2026 Official leaderboard
33 Gemini 3.8 Flash 1,583 high Sep 2026 Official leaderboard
34 MiniMax-M3.1-Flash 1,582 Oct 2026 Official leaderboard
35 DeepSeek-V4-Pro 1,582 high Apr 2026 Official leaderboard
36 DeepSeek-V4-Flash 1,582 high Apr 2026 Official leaderboard
37 GPT-6 Luna 1,579 max Sep 2026 Official leaderboard
38 Step 5 Preview 1,571 high Sep 2026 Official leaderboard
39 Claude Opus 4.7 1,558 high Apr 2026 Official leaderboard
40 Claude Opus 4.8 1,556 high May 2026 Official leaderboard
41 Grok 4.5 1,553 Jul 2026 Official leaderboard
42 Claude Opus 4.6 1,546 high Feb 2026 Official leaderboard
43 Mistral Large 4 1,544 Oct 2026 Official leaderboard
44 Muse Spark 1.1 1,542 Jul 2026 Official leaderboard
45 Claude Sonnet 5 1,540 high Jun 2026 Official leaderboard
46 Gemini 3.6 Flash 1,538 high Jul 2026 Official leaderboard
47 Muse Spark 1.2 1,534 xhigh Aug 2026 Official leaderboard
48 Muse Spark Open 1,526 max Oct 2026 Official leaderboard
49 Claude Sonnet 4.6 1,523 Feb 2026 Official leaderboard
50 GPT-5.6 Terra 1,520 xHigh, Codex Jul 2026 Official leaderboard
51 Seed2.1 Pro 1,519 Jun 2026 Official leaderboard
52 GPT-5.6 Luna 1,519 xHigh, Codex Jul 2026 Official leaderboard
53 Qwen3.7-Max 1,515 May 2026 Official leaderboard
54 GPT-5.5 1,512 xHigh, Codex Apr 2026 Official leaderboard
55 Kimi K2.6 1,509 Apr 2026 Official leaderboard
56 GLM-5.1 1,508 Apr 2026 Official leaderboard
57 Tencent Hy3 1,505 Jul 2026 Official leaderboard
58 Gemini 3.5 Flash 1,499 high May 2026 Official leaderboard
59 Claude Opus 4.5 1,494 high, 32k Nov 2025 Official leaderboard
60 MiniMax-M3 1,483 Jun 2026 Official leaderboard
61 Spark-X2.6 1,482 Oct 2026 Official leaderboard
62 Qwen 3.6 Max (Preview) 1,482 Apr 2026 Official leaderboard
63 MiMo-V2.5-Pro 1,480 Apr 2026 Official leaderboard
64 Kimi K2.7 Code 1,474 Jun 2026 Official leaderboard
65 GPT-5.4 1,463 High, Codex Mar 2026 Official leaderboard
66 Qwen 3.6 Plus 1,462 Mar 2026 Official leaderboard
67 Gemini 3.1 Pro 1,447 Feb 2026 Official leaderboard
68 Gemini 3 Flash 1,440 Dec 2025 Official leaderboard
69 Gemini 3 Pro 1,440 Nov 2025 Official leaderboard
70 MiMo-V2.5 1,438 Apr 2026 Official leaderboard
71 Kimi K2.5 1,436 thinking Jan 2026 Official leaderboard
72 Beam 1.0 1,436 Oct 2026 Official leaderboard
73 GLM-5 1,436 Feb 2026 Official leaderboard
74 GLM-4.7 1,435 Dec 2025 Official leaderboard
75 MiMo-V2-Pro 1,434 Mar 2026 Official leaderboard
76 Gemini 3.5 Flash-Lite 1,434 Jul 2026 Official leaderboard
77 GPT-5 1,417 medium Aug 2025 Official leaderboard
78 GPT-5.2 1,416 Dec 2025 Official leaderboard
79 Inkling 1,415 Jul 2026 Official leaderboard
80 Inkling-Small 1,409 Jul 2026 Official leaderboard
81 GPT-5.3 Codex 1,407 Codex Feb 2026 Official leaderboard
82 GLM-5V-Turbo 1,401 Apr 2026 Official leaderboard
83 Qwen3.5 397B-A17B 1,400 Feb 2026 Official leaderboard
84 Nemotron 3.5 Super VL 1,399 nvfp4 Oct 2026 Official leaderboard
85 MiniMax-M2.7 1,398 Mar 2026 Official leaderboard
86 GPT-5.4 Mini 1,397 high Mar 2026 Official leaderboard
87 GPT-5.1 1,396 medium Nov 2025 Official leaderboard
88 Claude Sonnet 4.5 1,394 high, 32k Sep 2025 Official leaderboard
89 Claude Opus 4.1 1,390 Aug 2025 Official leaderboard
90 MiniMax-M2.5 1,386 Feb 2026 Official leaderboard
91 MiniMax-M2.1 1,384 Dec 2025 Official leaderboard
92 Grok 4.20 Beta (Reasoning, 0309) 1,375 Oct 2026 Official leaderboard
93 Solar Pro 4 1,370 Aug 2026 Official leaderboard
94 Gemma 4 31B IT 1,369 Apr 2026 Official leaderboard
95 DeepSeek-V3.2 1,363 thinking Dec 2025 Official leaderboard
96 Qwen3.5-122B-A10B 1,360 Feb 2026 Official leaderboard
97 Gemma 4 26B A4B 1,359 Apr 2026 Official leaderboard
98 Qwen3.5-27B 1,358 Feb 2026 Official leaderboard
99 Grok 4.3 1,358 Apr 2026 Official leaderboard
100 Tencent Hy3 preview 1,357 Apr 2026 Official leaderboard
101 Muse Glimmer 1,355 Aug 2026 Official leaderboard
102 Laguna M.1 1,349 Apr 2026 Official leaderboard
103 GLM-4.6 1,341 Sep 2025 Official leaderboard
104 GPT-5.2 Codex 1,339 Dec 2025 Official leaderboard
105 GPT-5.1-Codex 1,337 Nov 2025 Official leaderboard
106 MiMo-V2-Flash 1,331 non-thinking Dec 2025 Official leaderboard
107 Claude Haiku 4.5 1,330 Oct 2025 Official leaderboard
108 Kimi K2 Thinking Turbo 1,322 Oct 2026 Official leaderboard
109 Laguna XS.2 1,302 Apr 2026 Official leaderboard
110 MiniMax-M2 1,297 Oct 2025 Official leaderboard
111 Qwen3-Coder-480B-A35B 1,275 Jul 2025 Official leaderboard
112 DeepSeek-V3.2-Exp 1,271 Sep 2025 Official leaderboard
113 Mistral Medium 3.5 1,264 Apr 2026 Official leaderboard
114 Gemini 3.1 Flash-Lite 1,256 Mar 2026 Official leaderboard
115 KAT-Coder-Pro V1 1,256 Oct 2026 Official leaderboard
116 Qwen3.5-35B-A3B 1,254 Feb 2026 Official leaderboard
117 Qwen3.5-Flash 1,244 Feb 2026 Official leaderboard
118 GPT-5.1-Codex-mini 1,244 Nov 2025 Official leaderboard
119 Grok 4.1 Fast 1,242 reasoning Nov 2025 Official leaderboard
120 Trinity Large Thinking 1,238 Apr 2026 Official leaderboard
121 Mistral Large 3 1,233 Dec 2025 Official leaderboard
122 Gemini 2.5 Pro Jun 2025 1,226 Jun 2025 Official leaderboard
123 Grok 4.1 1,214 thinking Nov 2025 Official leaderboard
124 Devstral 2 1,197 Dec 2025 Official leaderboard
125 Granite 4.1 8B 1,192 Apr 2026 Official leaderboard
126 Mercury 2 1,170 Feb 2026 Official leaderboard
127 grok-code-fast-1 1,166 Aug 2025 Official leaderboard
128 Grok 4 Fast 1,159 reasoning Sep 2025 Official leaderboard
129 Devstral Medium 1,071 Jul 2025 Official leaderboard

Labs report scores for 3 models in their cards. We plot every source in the charts below.

More effort adds a median 36 elo

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Claude Sonnet 5.5
high xhigh
1,770
Claude Opus 5
high
1,691
Muse Spark 1.3
xhigh max
1,656
DeepSeek-V4-Pro
high high
1,582
Claude Opus 4.7
high
1,558
Claude Opus 4.8
high
1,556
Claude Opus 4.6
high
1,546
Gemini 3.5 Flash
medium
1,499
Claude Opus 4.5
high, 32k
1,494
Gemini 3 Flash
minimal
1,440
Kimi K2.5
instant
1,436
GPT-5.1
medium
1,396
ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 max 1,810
GPT-6 Astra max 1,773
Claude Sonnet 5.5 xhigh 1,770
Gemini 4 Carbon high 1,768
GPT-6.1 Sol max 1,751
Claude Fable 5.1 max 1,746
Muse Wonder Full max 1,725
Claude Sonnet 5.5 high 1,711
Kimi K3.1 Pro max 1,695
Claude Opus 5 max 1,691
Gemini 4 Argon high 1,679
GPT-6 Sol max 1,678
Qwen3.8 Max (0902) 1,676
GPT-6.1 Terra max 1,673
Qwen 3.8 Max 1,672
DeepSeek V4.1 Flash max 1,665
Claude Opus 5 high 1,658
Kimi K3 max 1,656
Muse Spark 1.3 max 1,656
Muse Wonder max 1,652
Grok 4.7 xhigh 1,637
MiMo-V2.6 Flash 1,637
Tencent Hy4 preview 1,631
MiMo-V2.6 Pro 1,625
Claude Fable 5 high 1,624
GLM-5.3 max 1,623
Muse Spark 1.3 xhigh 1,620
DeepSeek V4.1 Flash max 1,619
Grok 4.6 high 1,616
GPT-5.6 Sol xHigh, Codex 1,616
GLM-5.3-Flash 1,607
GLM-5.2 max 1,606
Qwen 3.8 27B 1,590
Gemini 3.7 Flash high 1,590
Claude Haiku 5.5 high 1,589
Gemini 3.8 Flash high 1,583
MiniMax-M3.1-Flash 1,582
DeepSeek-V4-Pro high 1,582
DeepSeek-V4-Flash high 1,582
GPT-6 Luna max 1,579
Step 5 Preview high 1,571
Claude Opus 4.7 high 1,558
Claude Opus 4.8 high 1,556
Claude Opus 4.7 1,555
Grok 4.5 1,553
Claude Opus 4.6 high 1,546
Mistral Large 4 1,544
Muse Spark 1.1 1,542
Claude Sonnet 5 high 1,540
Gemini 3.6 Flash high 1,538
Claude Opus 4.6 1,537
Claude Opus 4.8 1,536
Muse Spark 1.2 xhigh 1,534
Muse Spark Open max 1,526
Claude Sonnet 4.6 1,523
GPT-5.6 Terra xHigh, Codex 1,520
Seed2.1 Pro 1,519
GPT-5.6 Luna xHigh, Codex 1,519
Qwen3.7-Max 1,515
GPT-5.5 xHigh, Codex 1,512
Kimi K2.6 1,509
GLM-5.1 1,508
Tencent Hy3 1,505
Gemini 3.5 Flash high 1,499
Claude Opus 4.5 high, 32k 1,494
Gemini 3.5 Flash medium 1,491
GPT-5.5 High, Codex 1,487
MiniMax-M3 1,483
Qwen 3.6 Max (Preview) 1,482
MiMo-V2.5-Pro 1,480
Kimi K2.7 Code 1,474
Claude Opus 4.5 1,469
DeepSeek-V4-Pro high 1,464
GPT-5.4 High, Codex 1,463
Qwen 3.6 Plus 1,462
GPT-5.5 Codex 1,456
DeepSeek-V4-Pro 1,447
Gemini 3.1 Pro 1,447
GPT-5.4 Medium, Codex 1,444
Gemini 3 Flash 1,440
Gemini 3 Pro 1,440
MiMo-V2.5 1,438
Kimi K2.5 thinking 1,436
Beam 1.0 1,436
GLM-5 1,436
GLM-4.7 1,435
MiMo-V2-Pro 1,434
Gemini 3.5 Flash-Lite 1,434
DeepSeek-V4-Flash high 1,430
GPT-5 medium 1,417
GPT-5.2 1,416
Inkling 1,415
Inkling-Small 1,409
GPT-5.3 Codex Codex 1,407
Kimi K2.5 instant 1,406
GPT-5.4 1,401
GLM-5V-Turbo 1,401
Qwen3.5 397B-A17B 1,400
Nemotron 3.5 Super VL nvfp4 1,399
MiniMax-M2.7 1,398
GPT-5.4 Mini high 1,397
GPT-5.1 medium 1,396
Claude Sonnet 4.5 high, 32k 1,394
Claude Opus 4.1 1,390
MiniMax-M2.5 1,386
Claude Sonnet 4.5 1,385
Gemini 3 Flash minimal 1,384
MiniMax-M2.1 1,384
GPT-5.3 Codex Codex 1,371
Solar Pro 4 1,370
Gemma 4 31B IT 1,369
DeepSeek-V3.2 thinking 1,363
Qwen3.5-122B-A10B 1,360
Gemma 4 26B A4B 1,359
Qwen3.5-27B 1,358
Grok 4.3 1,358
Tencent Hy3 preview 1,357
Muse Glimmer 1,355
Laguna M.1 1,349
GPT-5.1 1,341
GLM-4.6 1,341
GPT-5.2 Codex 1,339
GPT-5.1-Codex 1,337
MiMo-V2-Flash non-thinking 1,331
Claude Haiku 4.5 1,330
DeepSeek-V3.2 1,326
Laguna XS.2 1,302
MiniMax-M2 1,297
MiMo-V2-Flash thinking 1,292
Qwen3-Coder-480B-A35B 1,275
DeepSeek-V3.2-Exp 1,271
Mistral Medium 3.5 1,264
Gemini 3.1 Flash-Lite 1,256
Qwen3.5-35B-A3B 1,254
Qwen3.5-Flash 1,244
GPT-5.1-Codex-mini 1,244
Grok 4.1 Fast reasoning 1,242
Trinity Large Thinking 1,238
Mistral Large 3 1,233
Gemini 2.5 Pro Jun 2025 1,226
Grok 4.1 thinking 1,214
Devstral 2 1,197
Granite 4.1 8B 1,192
Mercury 2 1,170
grok-code-fast-1 1,166
Grok 4 Fast reasoning 1,159
Devstral Medium 1,071

12 models have held number one since its release in Nov 2025

Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.

OpenAI Google DeepMind Anthropic DeepSeek Moonshot AI Z.ai
Own leaderboard 12 leaders
Claude Opus 5.5 1,810
Lab claims Not tested since Aug 2026
Kimi K3 1,678

Scores over time

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

On the benchmark’s own leaderboard 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

One lab put it in a launch post in the last six months

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
13 Aug 2026 Gemini 3.7 Flash 1588 Code Arena (Web development, Elo)

3 system cards report it, from 3 labs

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
13 Aug 2026 Gemini 3.7 Flash model card 1588 Code Arena, web development, Elo (results table, p.5) Capabilities
27 Jul 2026 Kimi K3 technical report 1678 WebDev Arena (Elo), rank 1 of 99 (Table 5). Mapped to the current Arena WebDev board Other
17 Feb 2026 GLM-5 technical report 1449 LMArena Code Arena (Agentic WebDev) score, rank 6 (Figure 3 screenshot) Capabilities

Versions

Replaces WebDev 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