Benchmark Almanac

Code Migration v1.1

Claude Sonnet 5.5 86.9% leads since October 2026
Not in model cards

Models port a working program into another language offline and ship a Dockerfile. A hidden behavior-test suite the model never sees scores the port, after an anti-cheat check.

Same tasks as v1, but every model runs on the Pi agent harness and grading uses revised hidden test suites. Vals says scores are not directly comparable to v1. Part of Vals Index v2.2.

Made by
Vals AI
Released
October 2026
Kind
Fixing real repositories
Level
Expert
Scored
Checked automatically
Answers
Code
Tasks
Kept private
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last model added
November 2026
Scores refreshed
November 12th, 2026
30
Models scored on its own leaderboard

Claude Sonnet 5.5 leads with 86.9%

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.

30 models. Scroll sideways for the rest.

  1. 86.9 Claude Sonnet 5.5
  2. 83.3 Claude Fable 5.2
  3. 81.0 Claude Opus 5.5
  4. 80.6 GPT-6 Astra
  5. 77.3 GPT-6.1 Sol
  6. 72.4 Muse Wonder
  7. 70.8 Gemini 4 Argon
  8. 67.5 DeepSeek V4.1 Flash 1017
  9. 62.8 Muse Wonder Full
  10. 58.9 DeepSeek V4.1 Flash
  11. 57.2 Qwen3.8 Max (0902)
  12. 55.1 GLM-5.3
  13. 53.2 Grok 4.7
  14. 52.4 Claude Haiku 5.5
  15. 52.0 DeepSeek V4 Pro 0813
  16. 51.6 GLM-5.3-Flash
  17. 48.8 MiMo-V2.6 Pro
  18. 47.0 Step 5 Preview
  19. 46.7 GPT-6 Luna
  20. 45.4 Tencent Hy4 preview
  21. 39.5 MiMo-V2.6 Flash
  22. 39.3 Mercury 3
  23. 38.6 Gemini 3.8 Flash
  24. 28.6 Kimi K3
  25. 26.9 Muse Spark Open
  26. 22.9 Beam 1.0
  27. 20.9 MiniMax-M3
  28. 16.4 Nemotron 3.5 Super VL
  29. 10.4 Inkling-Small
  30. 8.82 Inkling
  1. 1 Claude Sonnet 5.5 86.9%
  2. 2 Claude Fable 5.2 83.3%
  3. 3 Claude Opus 5.5 81%
  4. 4 GPT-6 Astra 80.6%
  5. 5 GPT-6.1 Sol 77.3%
  6. 6 Muse Wonder 72.4%
  7. 7 Gemini 4 Argon 70.8%
  8. 8 DeepSeek V4.1 Flash 1017 67.5%
  9. 9 Muse Wonder Full 62.8%
  10. 10 DeepSeek V4.1 Flash 58.9%
  11. 11 Qwen3.8 Max (0902) 57.2%
  12. 12 GLM-5.3 55.1%
  13. 13 Grok 4.7 53.2%
  14. 14 Claude Haiku 5.5 52.4%
  15. 15 DeepSeek V4 Pro 0813 52%
  16. 16 GLM-5.3-Flash 51.6%
  17. 17 MiMo-V2.6 Pro 48.8%
  18. 18 Step 5 Preview 47%
  19. 19 GPT-6 Luna 46.7%
  20. 20 Tencent Hy4 preview 45.4%
  21. 21 MiMo-V2.6 Flash 39.5%
  22. 22 Mercury 3 39.3%
  23. 23 Gemini 3.8 Flash 38.6%
  24. 24 Kimi K3 28.6%
  25. 25 Muse Spark Open 26.9%
  26. 26 Beam 1.0 22.9%
  27. 27 MiniMax-M3 20.9%
  28. 28 Nemotron 3.5 Super VL 16.4%
  29. 29 Inkling-Small 10.4%
  30. 30 Inkling 8.82%

Source: Official Code Migration v1.1 leaderboard

#ModelScoreSettingDateSource
1 Claude Sonnet 5.5 86.9% Sep 2026 Official leaderboard
2 Claude Fable 5.2 83.3% Nov 2026 Official leaderboard
3 Claude Opus 5.5 81% Sep 2026 Official leaderboard
4 GPT-6 Astra 80.6% Sep 2026 Official leaderboard
5 GPT-6.1 Sol 77.3% Sep 2026 Official leaderboard
6 Muse Wonder 72.4% Oct 2026 Official leaderboard
7 Gemini 4 Argon 70.8% Sep 2026 Official leaderboard
8 DeepSeek V4.1 Flash 1017 67.5% Oct 2026 Official leaderboard
9 Muse Wonder Full 62.8% Oct 2026 Official leaderboard
10 DeepSeek V4.1 Flash 58.9% Sep 2026 Official leaderboard
11 Qwen3.8 Max (0902) 57.2% Sep 2026 Official leaderboard
12 GLM-5.3 55.1% Aug 2026 Official leaderboard
13 Grok 4.7 53.2% Sep 2026 Official leaderboard
14 Claude Haiku 5.5 52.4% Oct 2026 Official leaderboard
15 DeepSeek V4 Pro 0813 52% Aug 2026 Official leaderboard
16 GLM-5.3-Flash 51.6% Aug 2026 Official leaderboard
17 MiMo-V2.6 Pro 48.8% Sep 2026 Official leaderboard
18 Step 5 Preview 47% Sep 2026 Official leaderboard
19 GPT-6 Luna 46.7% Sep 2026 Official leaderboard
20 Tencent Hy4 preview 45.4% Aug 2026 Official leaderboard
21 MiMo-V2.6 Flash 39.5% Sep 2026 Official leaderboard
22 Mercury 3 39.3% Oct 2026 Official leaderboard
23 Gemini 3.8 Flash 38.6% Sep 2026 Official leaderboard
24 Kimi K3 28.6% Jul 2026 Official leaderboard
25 Muse Spark Open 26.9% Oct 2026 Official leaderboard
26 Beam 1.0 22.9% Oct 2026 Official leaderboard
27 MiniMax-M3 20.9% Jun 2026 Official leaderboard
28 Nemotron 3.5 Super VL 16.4% Oct 2026 Official leaderboard
29 Inkling-Small 10.4% Jul 2026 Official leaderboard
30 Inkling 8.82% Jul 2026 Official leaderboard

The best score, 86.9%, costs $44 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 OpenAI Anthropic Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Sonnet 5.5 GPT-6.1 Sol DeepSeek V4.1 Flash 1017 DeepSeek V4.1 Flash MiMo-V2.6 Pro MiMo-V2.6 Flash 0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Sonnet 5.5 GPT-6.1 Sol

Source: Official Code Migration v1.1 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Sonnet 5.5 86.9% $44
Claude Fable 5.2 83.3% $69
Claude Opus 5.5 81% $72
GPT-6 Astra 80.6% $62
GPT-6.1 Sol 77.3% $9.51
Muse Wonder 72.4% $48
Gemini 4 Argon 70.8% $45
DeepSeek V4.1 Flash 1017 67.5% $1.76
Muse Wonder Full 62.8% $24
DeepSeek V4.1 Flash 58.9% $1.09
Qwen3.8 Max (0902) 57.2% $27
GLM-5.3 55.1% $35
Grok 4.7 53.2% $38
Claude Haiku 5.5 52.4% $6.20
DeepSeek V4 Pro 0813 52% $12
GLM-5.3-Flash 51.6% $8.96
MiMo-V2.6 Pro 48.8% $0.99
Step 5 Preview 47% $8.61
GPT-6 Luna 46.7% $0.83
Tencent Hy4 preview 45.4% $3.58
MiMo-V2.6 Flash 39.5% $0.53
Mercury 3 39.3% $2.11
Gemini 3.8 Flash 38.6% $16
Kimi K3 28.6% $36
Muse Spark Open 26.9% $3.77
Beam 1.0 22.9%
MiniMax-M3 20.9% $8.39
Nemotron 3.5 Super VL 16.4% $1.60
Inkling-Small 10.4% $1.36
Inkling 8.82% $6.13

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 Oct 2026, by Claude Sonnet 5.5
Claude Sonnet 5.5 86.9%

Source: Official Code Migration v1.1 leaderboard

Claude Sonnet 5.5 on the official leaderboard passed 85% in September 2026

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 2027 Near the ceiling (85%) Claude Sonnet 5.5 Claude Opus 5.5 Qwen3.8 Max (0902) DeepSeek V4 Pro 0813 Kimi K3 MiniMax-M3 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Sonnet 5.5 GPT-6 Astra

Source: Official Code Migration v1.1 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.

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
Code Migration v1.1 Passed 85% in Oct 2026, by Claude Sonnet 5.5
Claude Sonnet 5.5 86.9%
Code Migration 5 leaders
Claude Sonnet 5.5 69.8%
Code Migration v1.1 Passed 85% in Oct 2026, by Claude Sonnet 5.5
Claude Sonnet 5.5 86.9%

Source: Official leaderboards

Replaces Code Migration

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