BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases12

SWE-bench Multimodal

GitHub issues from JavaScript repositories whose bugs are visual, so fixing them means reading images. Run by Vals AI.

As of 24 Oct 2026, Claude Opus 5.5 leads SWE-bench Multimodal on BenchLeader with 70.2%, ahead of Claude Fable 5.1 at 61.3%, across 20 model configurations with a published result.

Published by
Vals AI
Category
Coding
Index weight
Reference only
Models
20
Data as of
24 Oct 2026
Newest result
16 Oct 2026

Vals AI (vals.ai).

Top 15
  1. 1Claude Opus 5.570.2%
  2. 2Claude Fable 5.161.3%
  3. 3Claude Sonnet 5.558.3%
  4. 4Gemini 4 Argon (high)56.0%
  5. 5Muse Wonder Full (max)47.3%
  6. 6DeepSeek V4.1 Flash (max)46.5%
  7. 7Muse Wonder (max)45.6%
  8. 8Kimi K3 (max)44.2%
  9. 9Claude Haiku 5.544.0%
  10. 10GPT-6.1 Sol (max)43.5%
  11. 11GPT-6 Astra (max)43.3%
  12. 12Muse Spark 1.3 (max)43.1%
  13. 13Grok 4.7 (xhigh)42.9%
  14. 14Qwen3.8 Max (max)42.5%
  15. 15MiMo-V2.6-Pro41.3%
20 of 20
#
1Claude Opus 5.570.2%62.02026-10-16
2Claude Fable 5.161.3%66.42026-10-16
3Claude Sonnet 5.558.3%–2026-10-16
4Gemini 4 Argonhigh56.0%70.82026-10-16
5Muse Wonder Fullmax47.3%70.72026-10-16
6DeepSeek V4.1 Flashmax46.5%61.22026-10-16
7Muse Wondermax45.6%67.52026-10-16
8Kimi K3max44.2%64.72026-10-16
9Claude Haiku 5.544.0%–2026-10-16
10GPT-6.1 Solmax43.5%69.92026-10-16
11GPT-6 Astramax43.3%70.42026-10-16
12Muse Spark 1.3max43.1%67.12026-10-16
13Grok 4.7xhigh42.9%61.62026-10-16
14Qwen3.8 Maxmax42.5%63.32026-10-16
15MiMo-V2.6-Pro41.3%64.02026-10-16
16Gemini 3.8 Flashhigh40.8%64.42026-10-16
17Step 5 Preview40.8%60.42026-10-16
18GPT-6 Lunamax37.5%60.02026-10-16
19MiniMax M333.3%56.42026-10-16
20Mistral Large 4high33.1%49.82026-10-16

More coding benchmarks · All LLM benchmarks

Cite as: BenchLeader, “SWE-bench Multimodal leaderboard”, data as of 24 Oct 2026, https://www.benchleader.com/benchmarks/vals_swebench_multimodal.

SWE-bench Multimodal: questions

What does SWE-bench Multimodal measure?
GitHub issues from JavaScript repositories whose bugs are visual, so fixing them means reading images. Run by Vals AI. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads SWE-bench Multimodal?
Claude Opus 5.5 leads SWE-bench Multimodal with 70.2% as of 24 Oct 2026, ahead of Claude Fable 5.1 at 61.3%.
How many models have SWE-bench Multimodal results?
20 model configurations have a SWE-bench Multimodal result on BenchLeader, all taken from Vals AI.
Who runs SWE-bench Multimodal and how often is it updated?
SWE-bench Multimodal is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
Does SWE-bench Multimodal count toward the BenchLeader Index?
No. SWE-bench Multimodal is shown for reference but left out of the composite index.