Benchmark Almanac

SWE-Bench Pro V2

Claude Opus 5 99.4% leads since September 2026 Gemini 4 Argon 99.2% leads in mini-SWE-agent
Claude Opus 5 99.4% leads since September 2026
In use
Version 2.0

Agents resolve 642 repaired issue tasks from the SWE-Bench Pro public split in Go, Python, JavaScript and TypeScript. A locked protocol re-grades each patch in a clean container.

Audits and expert review repaired or retired 480 of the 731 v1 public tasks and dropped 89. The agent can reach only the model endpoint. V2 scores are not comparable with v1. The dataset default config is V2 since September 22nd, 2026.

Released
September 2026
Kind
Fixing real repositories
Level
Expert
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Code → Code, Actions
Interaction
Agentic
Environment
Terminal
Run by
Anyone
Based on
SWE-Bench Pro
In views
Agents
Strongest model tested
GPT-6 Astra #4 on the capability index
Last in a launch post
October 2026
Last model added
November 2026
Scores refreshed
November 20th, 2026

In the authors’ words

SWE-Bench Pro is a widely used benchmark of long-horizon software engineering tasks for coding agents. As agents improve, they reach task issues earlier agents never touched and interact with the evaluation environment in ways the original harness did not anticipate. We present SWE-Bench Pro V2, a revised public split built to keep the benchmark accurate as capabilities advance. Agentic audits, expert adjudication and blind re-implementation repaired or retired 480 of the 731 tasks: 23 contracted engineers spent 404 hours, each working in a language they write professionally, and every revised instruction was verified by a second engineer solving the task from the text alone. A hardened protocol restricts network access to the model endpoint and re-grades every patch in a clean container, so agents cannot retrieve upstream fixes or get credit for changes outside their patch. On the 642 V2 tasks frontier agents resolve up to 99.4% while other systems stay clearly separated, and a 51-task Hard split spreads systems from 52.9% to 98.0%. We release the benchmark and distill lessons for maintaining agent benchmarks over time.

Abstract of SWE-Bench Pro V2: A Cleaner, Harder-to-Game Coding Benchmark (arXiv:2610.24206, CC0)
18
Models scored on its own leaderboard
1
Launch post lists it, from 1 lab

Claude Opus 5 leads with 99.4% Claude Opus 5 leads with 99.4% Gemini 4 Argon leads with 99.2% in mini-SWE-agent

Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. This view keeps only runs in mini-SWE-agent, a harness no model’s maker built. A maker’s own agent, such as Claude Code with a Claude model, is left out.

Claude Opus 5 leads with 99.4%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.

  1. 1 Claude Opus 5 xhigh 99.4%
  2. 2 Gemini 4 Argon high 99.2%
  3. 3 Claude Fable 5.1 high 99.1%
  4. 4 Kimi K3 max 97.7%
  5. 5 Inkling 2 Small 97.7%
  6. 6 GPT-6 Astra high 96.9%
  7. 7 Muse Wonder xhigh 96.7%
  8. 8 MiniMax-M3.1-Flash xhigh 96.6%
  9. 9 GPT-6.1 Sol xhigh 96.1%
  10. 10 Claude Sonnet 5.5 xhigh 95.8%
  11. 11 GLM-5.3 max 95.6%
  12. 12 GPT-5.6 Sol xhigh 95.5%
  13. 13 Muse Spark Open xhigh 95%
  14. 14 Gemini 3.8 Flash high 94.9%
  15. 15 Claude Sonnet 5 xhigh 93.2%
  16. 16 GPT-5.6 Terra xhigh 92.4%
  17. 17 Inkling xhigh 89.9%
  18. 18 Claude Haiku 5.5 high 87.2%

Source: Official SWE-Bench Pro V2 leaderboard

#ModelScore ±SettingDateSource
1 Claude Opus 5 99.4% ±0.4 xhigh, Claude Code Jul 2026 Official leaderboard
2 Gemini 4 Argon 99.2% ±0.5 high, mini-SWE-agent Sep 2026 Official leaderboard
3 Claude Fable 5.1 99.1% ±0.5 high, Claude Code Sep 2026 Official leaderboard
4 Kimi K3 97.7% ±0.9 max, mini-SWE-agent Jul 2026 Official leaderboard
5 Inkling 2 Small 97.7% ±1.5 Nov 2026 Official leaderboard
6 GPT-6 Astra 96.9% ±1.1 high, Codex Sep 2026 Official leaderboard
7 Muse Wonder 96.7% ±1.1 xhigh, mini-SWE-agent Oct 2026 Official leaderboard
8 MiniMax-M3.1-Flash 96.6% ±1.2 xhigh, mini-SWE-agent Oct 2026 Official leaderboard
9 GPT-6.1 Sol 96.1% ±1.2 xhigh, mini-SWE-agent Sep 2026 Official leaderboard
10 Claude Sonnet 5.5 95.8% ±1.3 xhigh, mini-SWE-agent Sep 2026 Official leaderboard
11 GLM-5.3 95.6% ±1.3 max, mini-SWE-agent Aug 2026 Official leaderboard
12 GPT-5.6 Sol 95.5% ±1.4 xhigh, Codex Jul 2026 Official leaderboard
13 Muse Spark Open 95% ±1.4 xhigh, mini-SWE-agent Oct 2026 Official leaderboard
14 Gemini 3.8 Flash 94.9% ±1.5 high, mini-SWE-agent Sep 2026 Official leaderboard
15 Claude Sonnet 5 93.2% ±1.7 xhigh, Claude Code Jun 2026 Official leaderboard
16 GPT-5.6 Terra 92.4% ±1.8 xhigh, Codex Jul 2026 Official leaderboard
17 Inkling 89.9% ±2.1 xhigh, mini-SWE-agent Jul 2026 Official leaderboard
18 Claude Haiku 5.5 87.2% ±2.4 high, mini-SWE-agent Oct 2026 Official leaderboard
#ModelScore ±SettingDateSource
1 Gemini 4 Argon 99.2% ±0.5 high, mini-SWE-agent Sep 2026 Official leaderboard
2 Kimi K3 97.7% ±0.9 max, mini-SWE-agent Jul 2026 Official leaderboard
3 Inkling 2 Small 97.7% ±1.5 Nov 2026 Official leaderboard
4 Muse Wonder 96.7% ±1.1 xhigh, mini-SWE-agent Oct 2026 Official leaderboard
5 MiniMax-M3.1-Flash 96.6% ±1.2 xhigh, mini-SWE-agent Oct 2026 Official leaderboard
6 GPT-6.1 Sol 96.1% ±1.2 xhigh, mini-SWE-agent Sep 2026 Official leaderboard
7 Claude Sonnet 5.5 95.8% ±1.3 xhigh, mini-SWE-agent Sep 2026 Official leaderboard
8 GLM-5.3 95.6% ±1.3 max, mini-SWE-agent Aug 2026 Official leaderboard
9 Muse Spark Open 95% ±1.4 xhigh, mini-SWE-agent Oct 2026 Official leaderboard
10 Gemini 3.8 Flash 94.9% ±1.5 high, mini-SWE-agent Sep 2026 Official leaderboard
11 Inkling 89.9% ±2.1 xhigh, mini-SWE-agent Jul 2026 Official leaderboard
12 Claude Haiku 5.5 87.2% ±2.4 high, mini-SWE-agent Oct 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Opus 5 99.4% xhigh, Claude Code Jul 2026 Official leaderboard
2 Gemini 4 Argon 99.2% high, mini-SWE-agent Sep 2026 Official leaderboard
3 Claude Fable 5.1 99.1% high, Claude Code Sep 2026 Official leaderboard
4 Kimi K3 97.7% max, mini-SWE-agent Jul 2026 Official leaderboard
5 Inkling 2 Small 97.7% Nov 2026 Official leaderboard
6 GPT-6 Astra 96.9% high, Codex Sep 2026 Official leaderboard
7 Muse Wonder 96.7% xhigh, mini-SWE-agent Oct 2026 Official leaderboard
8 MiniMax-M3.1-Flash 96.6% xhigh, mini-SWE-agent Oct 2026 Official leaderboard
9 GPT-6.1 Sol 96.1% xhigh, mini-SWE-agent Sep 2026 Official leaderboard
10 Claude Sonnet 5.5 95.8% xhigh, mini-SWE-agent Sep 2026 Official leaderboard
11 GLM-5.3 95.6% max, mini-SWE-agent Aug 2026 Official leaderboard
12 GPT-5.6 Sol 95.5% xhigh, Codex Jul 2026 Official leaderboard
13 Muse Spark Open 95% xhigh, mini-SWE-agent Oct 2026 Official leaderboard
14 Gemini 3.8 Flash 94.9% high, mini-SWE-agent Sep 2026 Official leaderboard
15 Claude Sonnet 5 93.2% xhigh, Claude Code Jun 2026 Official leaderboard
16 GPT-5.6 Terra 92.4% xhigh, Codex Jul 2026 Official leaderboard
17 Inkling 89.9% xhigh, mini-SWE-agent Jul 2026 Official leaderboard
18 Flint 89.7% Oct 2026 Lab model card
19 Claude Haiku 5.5 87.2% high, mini-SWE-agent Oct 2026 Official leaderboard

Labs report scores for 1 model in their cards. Switch the scores above to rank them together.

Number one over time Number one over time

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

Anthropic
Own leaderboard Passed 85% in Sep 2026, by Claude Opus 5

Source: Official SWE-Bench Pro V2 leaderboard

Claude Sonnet 5 on the official leaderboard passed 85% in June 2026

Models released before it came out already scored that high. 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 Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Opus 5 Claude Sonnet 5 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Opus 5 Claude Sonnet 5

Sources: Official SWE-Bench Pro V2 leaderboard and labs’ model cards

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon Kimi K3 Inkling 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon Inkling

Sources: Official SWE-Bench Pro V2 leaderboard and labs’ model cards

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.

Recursive Superintelligence
DateLaunch postScoreAs the post writes it
Oct 21, 2026 Flint 89.7 SWE-bench Pro v2

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

Anthropic OpenAI Meta
SWE-Bench Pro V2 Passed 85% in Sep 2026, by Claude Opus 5
SWE-Bench Pro Replaced by V2, Sep 2026
SWE-Bench Pro V2 Passed 85% in Sep 2026, by Claude Opus 5

Source: Official leaderboards

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

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (18 entries) and launch posts (1 score).
  • Pass Models scored 19 models we track.
  • Pass Newest score The newest is from November 2026, for Inkling 2 Small.
  • Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Note Headroom Saturated. The best is 99.4%, past 85% since June 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on labs.scale.com: 18 rows for models we track at the last read.
  • Pass Last read November 20th, 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