Benchmark Almanac

Senior SWE-Bench

Claude Opus 5.5 39.3% leads since September 2026
Not in model cards

Agents build features and fix runtime bugs in real open-source repos from loose, natural instructions. Behavioral tests and codebase quality metrics give a tasteful solve rate.

100 tasks, 50 public. The leaderboard lists 23 models and reports tasteful solves. Launched June 30th, 2026.

Made by
Snorkel AI
Released
June 2026
Kind
Fixing real repositories
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Public, plus a private test set
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
23
Models scored on its own leaderboard

Claude Opus 5.5 leads with 39.3%

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

23 models. Scroll sideways for the rest.

  1. 39.3 Claude Opus 5.5
  2. 36.8 GPT-6 Astra
  3. 34.7 Claude Fable 5
  4. 34.7 GPT-5.6 Sol
  5. 34.7 Claude Opus 5
  6. 34.7 Claude Fable 5.1
  7. 30.5 Claude Opus 4.8
  8. 29.5 GPT-5.5
  9. 28.0 Claude Sonnet 5
  10. 27.7 Gemini 4 Argon
  11. 27.4 GPT-5.6 Terra
  12. 27.4 Grok 4.7
  13. 26.3 Grok 4.6
  14. 23.2 Grok 4.5
  15. 22.7 DeepSeek V4.1 Flash
  16. 21.1 MiniMax-M3
  17. 20.2 Kimi K3
  18. 18.1 Claude Opus 4.7
  19. 17.9 GLM-5.2
  20. 16.8 GPT-5.4
  21. 14.7 Gemini 3.7 Flash
  22. 6.32 Gemini 3.5 Flash
  23. 2.11 Gemini 3.1 Pro
  1. 1 Claude Opus 5.5 high 39.3%
  2. 2 GPT-6 Astra high 36.8%
  3. 3 Claude Fable 5 high 34.7%
  4. 4 GPT-5.6 Sol xhigh 34.7%
  5. 5 Claude Opus 5 high 34.7%
  6. 6 Claude Fable 5.1 medium 34.7%
  7. 7 Claude Opus 4.8 max 30.5%
  8. 8 GPT-5.5 xhigh 29.5%
  9. 9 Claude Sonnet 5 max 28%
  10. 10 Gemini 4 Argon medium 27.7%
  11. 11 GPT-5.6 Terra high 27.4%
  12. 12 Grok 4.7 xhigh 27.4%
  13. 13 Grok 4.6 high 26.3%
  14. 14 Grok 4.5 medium 23.2%
  15. 15 DeepSeek V4.1 Flash high 22.7%
  16. 16 MiniMax-M3 adaptive 21.1%
  17. 17 Kimi K3 max 20.2%
  18. 18 Claude Opus 4.7 max 18.1%
  19. 19 GLM-5.2 max 17.9%
  20. 20 GPT-5.4 xhigh 16.8%
  21. 21 Gemini 3.7 Flash high 14.7%
  22. 22 Gemini 3.5 Flash high 6.32%
  23. 23 Gemini 3.1 Pro high 2.11%

Source: Official Senior SWE-Bench leaderboard

#ModelScoreSettingDateSource
1 Claude Opus 5.5 39.3% high, mini-swe-agent Sep 2026 Official leaderboard
2 GPT-6 Astra 36.8% high, mini-swe-agent Sep 2026 Official leaderboard
3 Claude Fable 5 34.7% high, mini-swe-agent Jun 2026 Official leaderboard
4 GPT-5.6 Sol 34.7% xhigh, mini-swe-agent Jul 2026 Official leaderboard
5 Claude Opus 5 34.7% high, mini-swe-agent Jul 2026 Official leaderboard
6 Claude Fable 5.1 34.7% medium, mini-swe-agent Sep 2026 Official leaderboard
7 Claude Opus 4.8 30.5% max, mini-swe-agent May 2026 Official leaderboard
8 GPT-5.5 29.5% xhigh, mini-swe-agent Apr 2026 Official leaderboard
9 Claude Sonnet 5 28% max, mini-swe-agent Jun 2026 Official leaderboard
10 Gemini 4 Argon 27.7% medium, mini-swe-agent Sep 2026 Official leaderboard
11 GPT-5.6 Terra 27.4% high, mini-swe-agent Jul 2026 Official leaderboard
12 Grok 4.7 27.4% xhigh, mini-swe-agent Sep 2026 Official leaderboard
13 Grok 4.6 26.3% high, mini-swe-agent Aug 2026 Official leaderboard
14 Grok 4.5 23.2% medium, mini-swe-agent Jul 2026 Official leaderboard
15 DeepSeek V4.1 Flash 22.7% high, mini-swe-agent Sep 2026 Official leaderboard
16 MiniMax-M3 21.1% adaptive, mini-swe-agent Jun 2026 Official leaderboard
17 Kimi K3 20.2% max, mini-swe-agent Jul 2026 Official leaderboard
18 Claude Opus 4.7 18.1% max, mini-swe-agent Apr 2026 Official leaderboard
19 GLM-5.2 17.9% max, mini-swe-agent Jun 2026 Official leaderboard
20 GPT-5.4 16.8% xhigh, mini-swe-agent Mar 2026 Official leaderboard
21 Gemini 3.7 Flash 14.7% high, mini-swe-agent Aug 2026 Official leaderboard
22 Gemini 3.5 Flash 6.32% high, mini-swe-agent May 2026 Official leaderboard
23 Gemini 3.1 Pro 2.11% high, mini-swe-agent Feb 2026 Official leaderboard

The best score, 39.3%, costs $3.07 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Z.ai OpenAI Anthropic Cost frontier One model at each effort level
0% 10% 20% 30% 40% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Opus 5.5, high GPT-6 Astra, high GPT-5.6 Terra, high GLM-5.2, max 0% 10% 20% 30% 40% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Opus 5.5, high GPT-6 Astra, high

Source: Official Senior SWE-Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 high mini-swe-agent 39.3% $3.07 41,259
GPT-6 Astra high mini-swe-agent 36.8% $2.32 19,036
Claude Fable 5 high mini-swe-agent 34.7% $14 58,383
GPT-5.6 Sol xhigh mini-swe-agent 34.7% $3.43 32,661
Claude Opus 5 high mini-swe-agent 34.7% $9.28 70,959
Claude Fable 5.1 medium mini-swe-agent 34.7% $7.14 37,861
Claude Opus 4.8 max mini-swe-agent 30.5% $12 134,219
GPT-5.5 xhigh mini-swe-agent 29.5% $4.86 36,124
Claude Sonnet 5 max mini-swe-agent 28% $23 180,720
Gemini 4 Argon medium mini-swe-agent 27.7% $4.96 29,313
GPT-5.6 Terra high mini-swe-agent 27.4% $0.66 18,213
Grok 4.7 xhigh mini-swe-agent 27.4% $3.85 39,449
Grok 4.6 high mini-swe-agent 26.3% $2.26 23,924
Grok 4.5 medium mini-swe-agent 23.2% $1.01 21,222
DeepSeek V4.1 Flash high mini-swe-agent 22.7% $0.74 441,524
MiniMax-M3 adaptive mini-swe-agent 21.1% $1.27 75,787
Kimi K3 max mini-swe-agent 20.2% $7.03 54,819
Claude Opus 4.7 max mini-swe-agent 18.1% $22 100,938
GLM-5.2 max mini-swe-agent 17.9% $0.64 66,230
GPT-5.4 xhigh mini-swe-agent 16.8% $3.07 53,854
Gemini 3.7 Flash high mini-swe-agent 14.7% $2.75 47,266
Gemini 3.5 Flash high mini-swe-agent 6.32% $4.37 70,373
Gemini 3.1 Pro high mini-swe-agent 2.11% $1.33 17,278

4 models have held number one on the official leaderboard since its release in June 2026

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

Anthropic OpenAI
Own leaderboard 4 leaders
Claude Opus 5.5 39.3%

Source: Official Senior SWE-Bench leaderboard

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 Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5 Claude Opus 4.8 Claude Opus 4.7 Gemini 3.1 Pro 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5.5 Claude Opus 4.8

Source: Official Senior SWE-Bench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Coding

All 282 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