Benchmark Almanac

SWE-bench Multimodal v2

Claude Opus 5.5 70.2% leads since September 2026
Claude Opus 5.5 70.2% leads since September 2026
Only in system cards
Version 2

The official leaderboard hasn’t scored a model released after Claude Fable 5.2 (November 2026). Labs report their own numbers for one newer model: Inkling 2 Mini at 37.9%.

Agents patch JavaScript repositories using issue reports that include images. Automated tests check whether each patch fixes the issue.

The authors released 480 reproducible test tasks and rebuilt the grading environments. On November 9, 2026, Claude Opus 5.5 led the Vals AI board at 70.21%, ahead of Claude Fable 5.2 and 5.1. In a separate 480-task run, GPT-6 Luna passed 36.9% with images and 37.1% without them.

Made by
SWE-bench team
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, Images → Code, Actions
Interaction
Agentic
Environment
Terminal
Items from
Real work and use
Run by
Anyone
In views
Agents, Vision
Strongest model tested
Claude Fable 5.2 #2 on the capability index
Last in a system card
November 2026
Last model added
November 2026
Scores refreshed
November 20th, 2026
22
Models scored on its own leaderboard
1
System card reports it, from 1 lab

Claude Opus 5.5 leads with 70.2%

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. Every run is in the next chart.

Claude Opus 5.5 leads with 70.2%, 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 5.5 70.2%
  2. 2 Claude Fable 5.2 62.3%
  3. 3 Claude Fable 5.1 61.3%
  4. 4 Claude Sonnet 5.5 58.3%
  5. 5 Gemini 4 Argon 56%
  6. 6 Muse Wonder Full 47.3%
  7. 7 DeepSeek V4.1 Flash 46.5%
  8. 8 Muse Wonder 45.6%
  9. 9 Kimi K3 44.2%
  10. 10 Claude Haiku 5.5 44%
  11. 11 GPT-6.1 Sol 43.5%
  12. 12 GPT-6 Astra 43.3%
  13. 13 Muse Spark 1.3 max 43.1%
  14. 14 Inkling 2 Small 43.1%
  15. 15 Grok 4.7 42.9%
  16. 16 Qwen 3.8 Max 42.5%
  17. 17 MiMo-V2.6 Pro 41.3%
  18. 18 Gemini 3.8 Flash 40.8%
  19. 19 Step 5 Preview 40.8%
  20. 20 GPT-6 Luna 37.5%
  21. 21 MiniMax-M3 33.3%
  22. 22 Mistral Large 4 Preview 33.1%

Source: Official SWE-bench Multimodal v2 leaderboard

#ModelScore ±SettingDateSource
1 Claude Opus 5.5 70.2% ±4.1 Sep 2026 Official leaderboard
2 Claude Fable 5.2 62.3% ±4.3 Nov 2026 Official leaderboard
3 Claude Fable 5.1 61.3% ±4.4 Sep 2026 Official leaderboard
4 Claude Sonnet 5.5 58.3% ±4.4 Sep 2026 Official leaderboard
5 Gemini 4 Argon 56% ±4.4 Sep 2026 Official leaderboard
6 Muse Wonder Full 47.3% ±4.5 Oct 2026 Official leaderboard
7 DeepSeek V4.1 Flash 46.5% ±4.5 Sep 2026 Official leaderboard
8 Muse Wonder 45.6% ±4.5 Oct 2026 Official leaderboard
9 Kimi K3 44.2% ±4.4 Jul 2026 Official leaderboard
10 Claude Haiku 5.5 44% ±4.4 Oct 2026 Official leaderboard
11 GPT-6.1 Sol 43.5% ±4.4 Sep 2026 Official leaderboard
12 GPT-6 Astra 43.3% ±4.4 Sep 2026 Official leaderboard
13 Muse Spark 1.3 43.1% ±4.4 max Sep 2026 Official leaderboard
14 Inkling 2 Small 43.1% ±4.4 Nov 2026 Official leaderboard
15 Grok 4.7 42.9% ±4.4 Sep 2026 Official leaderboard
16 Qwen 3.8 Max 42.5% ±4.4 Aug 2026 Official leaderboard
17 MiMo-V2.6 Pro 41.3% ±4.4 Sep 2026 Official leaderboard
18 Gemini 3.8 Flash 40.8% ±4.4 Sep 2026 Official leaderboard
19 Step 5 Preview 40.8% ±4.4 Sep 2026 Official leaderboard
20 GPT-6 Luna 37.5% ±4.3 Sep 2026 Official leaderboard
21 MiniMax-M3 33.3% ±4.2 Jun 2026 Official leaderboard
22 Mistral Large 4 Preview 33.1% ±4.2 Oct 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Opus 5.5 70.2% Sep 2026 Official leaderboard
2 Claude Fable 5.2 62.3% Nov 2026 Official leaderboard
3 Claude Fable 5.1 61.3% Sep 2026 Official leaderboard
4 Claude Sonnet 5.5 58.3% Sep 2026 Official leaderboard
5 Gemini 4 Argon 56% Sep 2026 Official leaderboard
6 Muse Wonder Full 47.3% Oct 2026 Official leaderboard
7 DeepSeek V4.1 Flash 46.5% Sep 2026 Official leaderboard
8 Muse Wonder 45.6% Oct 2026 Official leaderboard
9 Kimi K3 44.2% Jul 2026 Official leaderboard
10 Claude Haiku 5.5 44% Oct 2026 Official leaderboard
11 GPT-6.1 Sol 43.5% Sep 2026 Official leaderboard
12 GPT-6 Astra 43.3% Sep 2026 Official leaderboard
13 Muse Spark 1.3 43.1% max Sep 2026 Official leaderboard
14 Inkling 2 Small 43.1% Nov 2026 Official leaderboard
15 Grok 4.7 42.9% Sep 2026 Official leaderboard
16 Qwen 3.8 Max 42.5% Aug 2026 Official leaderboard
17 MiMo-V2.6 Pro 41.3% Sep 2026 Official leaderboard
18 Gemini 3.8 Flash 40.8% Sep 2026 Official leaderboard
19 Step 5 Preview 40.8% Sep 2026 Official leaderboard
20 Inkling 2 Mini 37.9% Nov 2026 Lab model card
21 GPT-6 Luna 37.5% Sep 2026 Official leaderboard
22 MiniMax-M3 33.3% Jun 2026 Official leaderboard
23 Mistral Large 4 Preview 33.1% Oct 2026 Official leaderboard

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

The best score, 70.2%, costs $3.63 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.

DeepSeek Xiaomi Corp Anthropic OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Claude Opus 5.5 Claude Fable 5.2 DeepSeek V4.1 Flash MiMo-V2.6 Pro GPT-6 Luna 0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Claude Opus 5.5 Claude Fable 5.2

Source: Official SWE-bench Multimodal v2 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 70.2% $3.63 79,324
Claude Fable 5.2 62.3% $1.84 25,322
Claude Fable 5.1 61.3% $5.46 58,216
Claude Sonnet 5.5 58.3% $3.26 140,591
Gemini 4 Argon 56% $11 117,302
Muse Wonder Full 47.3% $1.86 32,389
DeepSeek V4.1 Flash 46.5% $0.23 86,458
Muse Wonder 45.6% $1.70 29,688
Kimi K3 44.2% $1.65 38,809
Claude Haiku 5.5 44% $0.70 129,659
GPT-6.1 Sol 43.5% $0.35 18,232
GPT-6 Astra 43.3% $1.81 16,208
Muse Spark 1.3 max 43.1% $0.86 34,503
Inkling 2 Small 43.1% $0.30 52,488
Grok 4.7 42.9% $4.66 74,479
Qwen 3.8 Max 42.5% $0.99 39,700
MiMo-V2.6 Pro 41.3% $0.12 67,485
Gemini 3.8 Flash 40.8% $3.69 80,953
Step 5 Preview 40.8% $0.61 56,648
GPT-6 Luna 37.5% $0.03 27,262
MiniMax-M3 33.3% $1.02 34,365
Mistral Large 4 Preview 33.1% $2.64 72,672

2 models have held number one on the official leaderboard since its release in September 2026 2 models have held number one with claims included since its release in September 2026

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

Anthropic
Own leaderboard 2 leaders

Source: Official SWE-bench Multimodal v2 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 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 5.5 Claude Fable 5.1 Kimi K3 MiniMax-M3 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5.1

Sources: Official SWE-bench Multimodal v2 leaderboard and labs’ model cards

Labs report it only in system cards

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

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. Anthropic accounts for 10 of the 13 releases that named it since 2024.

DateDocumentScoreAs the document writes itSection
Nov 20, 2026 Inkling 2 Mini model card 37.9 SWEBench Multimodal v2 (mini-swe-agent, Vals) Capabilities

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
SWE-bench Multimodal v2 6 leaders

Sources: Official leaderboards and labs’ model cards

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 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (22 entries) and system cards (1 score).
  • Pass Models scored 23 models we track.
  • Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
  • Pass Frontier models 7 of the top 10 have a score, led by Claude Fable 5.2 (#2).
  • Pass Headroom The best is 70.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on vals.ai: 22 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