Vals Code Migration (Code Migration)

We show this table for reference; we do not rank on it.

Data verified September 27, 202636 confirmed releases in the last 30 daysFollow model changes

Can language models reimplement real-world programs in another language?

Code Migration score on Code Migration — September 27, 2026

We mirror the published code migration score view for Code Migration. Claude Sonnet 5.5 leads the public snapshot at 69.83%, followed by GPT-6 Astra (67.74%) and Claude Opus 5.5 (66.65%). We do not use these results to rank models overall.

1Proprietary

Claude Sonnet 5.5

Anthropic

69.83%

Overall 80.9Context 1M

2Proprietary

GPT-6 Astra

OpenAI

max reasoning

67.74%

Overall 88.22Context 1.05M

3Proprietary

Claude Opus 5.5

Anthropic

66.65%

Overall 86.94Context 1M

69 modelsCodingCurrentDisplay onlyUpdated September 27, 2026

Code Migration score table (69 models)

Score

Claude Sonnet 5.5Anthropic · Closed

69.83%

GPT-6 AstraOpenAI · Closedmax reasoning

67.74%

Claude Opus 5.5Anthropic · Closed

66.65%

Claude Opus 5Anthropic · Closed

57.47%

GPT-6 SolOpenAI · Closedmax reasoning

57.20%

Claude Fable 5Anthropic · Closed

55.06%

Claude Fable 5.1Anthropic · Closed

54.61%

GPT-5.6 SolOpenAI · Closedmax reasoning

52.92%

GPT-5.6 TerraOpenAI · Closedmax reasoning

47.80%

Hy4 previewTencent · Open weight

47.43%

Muse Spark 1.3 MaxMetamax reasoning

47.41%

Claude Opus 4.8Anthropic · Closed

47.25%

DeepSeek V4.1 FlashDeepSeek · Open weighthigh reasoning

45.62%

GPT-5.5OpenAI · Closedxhigh reasoning

45.16%

Grok 4.7xAI · Closedxhigh reasoning

44.82%

Grok 4.6xAI · Closedhigh reasoning

44.57%

GPT-5.6 LunaOpenAI · Closedmax reasoning

44.55%

Claude Sonnet 5Anthropic · Closed

44.39%

GLM-5.3Z.AI · Open weightmax reasoning

44.22%

Claude Opus 4.7Anthropic · Closed

43.88%

MiMo-V2.6-ProXiaomi · Open weight

43.01%

GPT-6 LunaOpenAI · Closedmax reasoning

42.55%

DeepSeek V4 Pro 0813DeepSeek · Open weightmax reasoning

41.54%

MiMo-V2.6-FlashXiaomi · Open weight

40.93%

Claude Sonnet 4.6Anthropic · Closed

39.89%

DeepSeek V4 Flash 0731DeepSeek · Open weighthigh reasoning

38.63%

GLM-5.2Z.AI · Open weightmax reasoning

37.87%

Grok 4.5xAI · Closedhigh reasoning

36.60%

Gemini 3.8 FlashGoogle · Closedhigh reasoning

36.55%

GPT-5.4OpenAI · Closedxhigh reasoning

34.98%

Gemini 3.7 FlashGoogle · Closedhigh reasoning

34.80%

Muse Spark 1.1Meta · Closedxhigh reasoning

31.11%

Gemini 3.6 FlashGoogle · Closedhigh reasoning

30.93%

Muse Spark 1.2Meta · Closedxhigh reasoning

29.95%

Kimi K2.6Moonshot AI · Open weight

27.77%

Muse Spark 1.3Meta · Closedxhigh reasoning

27.58%

Gemini 3.5 FlashGoogle · Closedhigh reasoning

26.75%

DeepSeek V4 Pro 0813DeepSeek · Open weightmax reasoning

26.20%

GLM-5.1Z.AI · Open weight

25.77%

Kimi K2.7 CodeMoonshot AI · Open weight

25.39%

Qwen3.8 MaxAlibaba · Open weight

23.96%

MiMo-V2.5-ProXiaomi · Closed

21.56%

GLM-5.3-FlashZ.AI · Open weightmax reasoning

20.52%

MiniMax M3MiniMax · Open weight

19.93%

Gemini 3.1 Pro PreviewGooglehigh reasoning

17.31%

Kimi K3Moonshot AI · Closedmax reasoning

16.10%

GPT-5.4 nanoOpenAI · Closedhigh reasoning

14.47%

MiMo-V2.5Xiaomi · Closed

14.23%

Qwen3.8-27BAlibaba · Open weightxhigh reasoning

14.16%

Inkling-SmallThinking Machines Lab · Open weight0.99 reasoning

13.68%

Qwen3.7 MaxAlibaba · Closed

13.07%

GPT-5.4 miniOpenAI · Closedxhigh reasoning

12.94%

Qwen3.7 PlusAlibaba · Closed

12.86%

InklingThinking Machines Lab · Open weight0.99 reasoning

11.79%

Qwen3.6 PlusAlibaba · Closed

11.10%

MiniMax M2.7MiniMax · Open weight

8.69%

Claude Haiku 4.5 ThinkingAnthropic · Closed

7.55%

Kimi K2.5 ThinkingMoonshot AI

6.96%

Grok 4.3xAI · Closedhigh reasoning

6.79%

Gemini 3 Flash PreviewGooglehigh reasoning

6.39%

Gemini 3.5 Flash-LiteGoogle · Closedhigh reasoning

6.15%

Mistral Medium 3.5Mistral AIhigh reasoning

5.13%

Nemotron 3 Ultra 550b A55bNvidia

4.91%

Gemini 3.1 Flash Lite PreviewGooglehigh reasoning

4.61%

Mercury 2.5Inception · Closedhigh reasoning

4.49%

Nemotron Lightning 3p5 30b A3bFireworks AI

2.95%

Ling 3.0 Flash Af Rc3Ant

1.82%

Grok 4.20 0309 ReasoningxAI

0.31%

Ling 3.0 Flash 2607Ant

0.00%

How Code Migration is shown here

BenchLM mirrors the public Vals AI Code Migration leaderboard captured from https://www.vals.ai/benchmarks/code-migration and updated by Vals on September 27, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

Code Migration is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

Snapshot

69 Vals rows4 task viewsprivate datasetTasks: Overall, CLI, COBOL, Code QualityDisplay only

Code Migration on Vals AI Vals methodology Vals home

The published Code Migration snapshot places Claude Sonnet 5.5 first at 69.83%. The third row is 3.18 points behind. The broader top-10 range is 22.39 points, so the table still separates the published systems.

69 models have been evaluated on Code Migration. The benchmark falls in the Coding category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. Code Migration is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Code Migration

Year

2026

Tasks

Real-world program reimplementation in another language

Format

Accuracy score

Difficulty

Production code migration

BenchLM mirrors the public Vals AI Code Migration leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.

Code Migration Public benchmark source

Freshness and provenance

Version

Code Migration 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does Code Migration measure?

Can language models reimplement real-world programs in another language?

Which model leads the published Code Migration snapshot?

Claude Sonnet 5.5 currently leads the published Code Migration snapshot with 69.83% code migration score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on Code Migration?

The September 27, 2026 snapshot contains 69 AI models.

Last updated: September 27, 2026 · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.