Benchmark Almanac

Windsurf Arena Mode Leaderboard

Claude Opus 4.6 1,098 leads since February 2026
Not in model cards

Developers compare two model responses to the same coding request inside the editor. Their choices determine model Elo ratings.

The final table ranks 13 models. Cognition closed the board on June 30, 2026.

Made by
Cognition
Released
February 2026
Kind
Fixing real repositories
Level
University
Scored
Ranked by votes
Answers
Agent task
Tasks
Kept private
Measures
Preference
Language
English
In and out
Text, Code → Code
Interaction
Agentic
Environment
Windsurf
Items from
Real work and use
Refresh
One fixed set
Run by
A host
In views
Agents
Strongest model tested
Claude Opus 4.6 #39 on the capability index
Last model added
April 2026
Scores refreshed
November 21st, 2026
15
Entries on its own leaderboard, 9 models we track

Claude Opus 4.6 leads with 1,098 among models released by April 2026 Claude Opus 4.6 leads with 1,098 among models released by April 2026

Best score per entry from the benchmark's own leaderboard, read on November 21st, 2026. The leaderboard ranks models by ELO rating from Arena Mode usage. 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. This view shows only models we track, each at its best.

The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

Bars start at 900.

  1. 1 Claude Opus 4.6 thinking 1,098
  2. 2 Claude Opus 4.5 1,081
  3. 3 Claude Sonnet 4.5 1,058
  4. 4 GPT-5.4 Low Thinking 1,031
  5. 5 Kimi K2.6 1,028
  6. 6 GPT-5.5 Low Thinking 1,022
  7. 7 Claude Haiku 4.5 1,015
  8. 8 GPT-5.2 Low Thinking 1,011
  9. 9 GPT-5.2 Medium Thinking 998
  10. 10 GPT-5.3 Codex low 959
  11. 11 Gemini 3 Flash low 953
  12. 12 Gemini 3.1 Pro Low Thinking 937
  13. 13 Gemini 3.1 Pro High Thinking 936
  14. 14 GPT-5.3-Codex-Spark medium 914
  15. 15 grok-code-fast-1 911

Source: Official Windsurf Arena Mode Leaderboard leaderboard

#EntryScore ±SettingDateSource
1 Claude Opus 4.6 1,098 1085 to 1113 thinking Feb 2026 Official leaderboard
2 Claude Opus 4.5 1,081 1061 to 1099 Nov 2025 Official leaderboard
3 Claude Sonnet 4.5 1,058 1038 to 1074 Sep 2025 Official leaderboard
4 GPT-5.4 Low Thinking 1,031 1007 to 1053 Official leaderboard
5 Kimi K2.6 1,028 ±65 Apr 2026 Official leaderboard
6 GPT-5.5 Low Thinking 1,022 975 to 1063 Official leaderboard
7 Claude Haiku 4.5 1,015 ±30 Oct 2025 Official leaderboard
8 GPT-5.2 Low Thinking 1,011 994 to 1029 Official leaderboard
9 GPT-5.2 Medium Thinking 998 ±19 Official leaderboard
10 GPT-5.3 Codex 959 ±26.5 low Feb 2026 Official leaderboard
11 Gemini 3 Flash 953 ±27.5 low Dec 2025 Official leaderboard
12 Gemini 3.1 Pro Low Thinking 937 908 to 964 Official leaderboard
13 Gemini 3.1 Pro High Thinking 936 ±27.5 Official leaderboard
14 GPT-5.3-Codex-Spark 914 ±34.5 medium Feb 2026 Official leaderboard
15 grok-code-fast-1 911 ±27.5 Aug 2025 Official leaderboard
#ModelScore ±SettingDateSource
1 Claude Opus 4.6 1,098 1085 to 1113 thinking Feb 2026 Official leaderboard
2 Claude Opus 4.5 1,081 1061 to 1099 Nov 2025 Official leaderboard
3 Claude Sonnet 4.5 1,058 1038 to 1074 Sep 2025 Official leaderboard
4 Kimi K2.6 1,028 ±65 Apr 2026 Official leaderboard
5 Claude Haiku 4.5 1,015 ±30 Oct 2025 Official leaderboard
6 GPT-5.3 Codex 959 ±26.5 low Feb 2026 Official leaderboard
7 Gemini 3 Flash 953 ±27.5 low Dec 2025 Official leaderboard
8 GPT-5.3-Codex-Spark 914 ±34.5 medium Feb 2026 Official leaderboard
9 grok-code-fast-1 911 ±27.5 Aug 2025 Official leaderboard

Scores at each effort level

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 Opus 4.6
thinking
1,098
Claude Opus 4.5
thinking
1,081
Claude Sonnet 4.5
thinking
1,058

Source: Official Windsurf Arena Mode Leaderboard leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 4.6 thinking 1,098
Claude Opus 4.6 1,096
Claude Opus 4.5 1,081
Claude Sonnet 4.5 1,058
Claude Opus 4.5 thinking 1,058
Claude Sonnet 4.5 thinking 1,049
Kimi K2.6 1,028
Claude Haiku 4.5 1,015
GPT-5.3 Codex low 959
Gemini 3 Flash low 953
GPT-5.3-Codex-Spark medium 914
grok-code-fast-1 911

Scores at each effort level

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 Opus 4.6
thinking
1,098
Claude Opus 4.5
thinking
1,081
Claude Sonnet 4.5
thinking
1,058

Source: Official Windsurf Arena Mode Leaderboard leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 4.6 thinking 1,098
Claude Opus 4.6 1,096
Claude Opus 4.5 1,081
Claude Sonnet 4.5 1,058
Claude Opus 4.5 thinking 1,058
Claude Sonnet 4.5 thinking 1,049
Kimi K2.6 1,028
Claude Haiku 4.5 1,015
GPT-5.3 Codex low 959
Gemini 3 Flash low 953
GPT-5.3-Codex-Spark medium 914
grok-code-fast-1 911

2 models have held number one on the official leaderboard since its release in February 2026

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

Anthropic
Own leaderboard Not tested since Apr 2026

Source: Official Windsurf Arena Mode Leaderboard leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard
900 950 1,000 1,050 1,100 2026 900 950 1,000 1,050 1,100 2026

Source: Official Windsurf Arena Mode Leaderboard leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Coding

All 523 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%
  • 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%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Fable 5.2 96.6%
  • 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

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for Kimi K2.6. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (15 entries).
  • Pass Models scored 9 models we track, plus 6 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for Kimi K2.6.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 4.6 (#39).
  • Note Headroom Scored as an Elo rating, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on devin.ai: 12 rows for models we track and 6 other entries at the last read.
  • Pass Last read November 21st, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check