Benchmark profile

Vals-hosted SWE-bench mirror (Vals SWE-bench mirror)

Vals AI hosted SWE-bench view for solving production software engineering tasks.

Data verified July 22, 2026

How BenchLM shows Vals SWE-bench mirror

BenchLM mirrors the public Vals AI Vals SWE-bench mirror leaderboard captured from https://www.vals.ai/benchmarks/swebench and updated by Vals on July 22, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

Vals SWE-bench mirror 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.

75 Vals rows5 task viewspublic datasetTasks: Overall, 1-4 hours, 15 min - 1 hour, <15 min fix, >4 hoursDisplay only

Vals SWE-bench mirror on Vals AI Vals methodology Vals home

Vals SWE-bench score on Vals SWE-bench mirror — July 22, 2026

BenchLM mirrors the published vals swe-bench score view for Vals SWE-bench mirror. Claude Opus 5 leads the public snapshot at 97.00% , followed by GPT-5.6 Sol (96.20%) and Claude Fable 5 (95.00%). BenchLM does not use these results to rank models overall.

Claude Opus 5

Anthropic

anthropic/claude-opus-5

97.00%

Overall —

GPT-5.6 Sol

OpenAI

openai/gpt-5.6-sol

96.20%

Overall —

Claude Fable 5

Anthropic

anthropic/claude-fable-5

95.00%

Overall —

75 modelsExternal benchmark mirrorsCurrentDisplay onlyUpdated July 22, 2026

Vals SWE-bench score table (75 models)

Score

Claude Opus 5Anthropic

97.00%

GPT-5.6 SolOpenAI

96.20%

Claude Fable 5Anthropic

95.00%

Kimi K3Moonshot AI

93.40%

GPT-5.6 LunaOpenAI

93.00%

Claude Opus 4.8Anthropic

88.60%

Grok 4.5xAI

86.60%

Claude Opus 4.8 Claude CodeAnthropicClaude Code

85.80%

GLM 5.2Zhipu AI

82.80%

GPT-5.5OpenAI

82.60%

Muse Spark 1.1Meta

82.00%

Claude Opus 4.7Anthropic

82.00%

Composer 2.5CursorCursor CLI

79.60%

Gemini 3.6 FlashGoogle

79.60%

Claude Sonnet 5Anthropic

79.60%

Gemini 3.5 FlashGoogle

78.80%

Gemini 3.1 Pro PreviewGoogle

78.80%

GPT-5.4OpenAI

78.20%

Claude Opus 4.6 ThinkingAnthropic

78.20%

Kimi K2.7 CodeMoonshot AI

78.20%

GPT-5.3 CodexOpenAI

78.00%

InklingThinkingmachines

77.60%

Claude Sonnet 4.6Anthropic

77.40%

DeepSeek V4 ProDeepSeek

77.40%

GPT-5.5 CodexOpenAICodex

76.40%

Claude Opus 4.5 20251101 ThinkingAnthropic

76.40%

Gemini 3 Pro PreviewGoogle

76.40%

GLM 5.1Zhipu AI

76.40%

GPT-5.5 FactoryOpenAIFactory

76.20%

Kimi K2.6Moonshot AI

76.20%

GPT-5.2OpenAI

75.80%

GPT-5.6 TerraOpenAI

75.20%

Gemini 3 Flash PreviewGoogle

75.00%

Gemini 3.5 Flash LiteGoogle

75.00%

MiniMax M3MiniMax

75.00%

MiniMax M2.1MiniMax

74.80%

Muse SparkMeta

74.40%

MiniMax M2.5MiniMax

74.20%

Mimo V2.5 ProXiaomi

74.00%

MiniMax M2.7MiniMax

73.80%

Qwen3.6 PlusAlibaba

73.40%

GPT-5.4 MiniOpenAI

73.00%

Qwen3.6 Max PreviewAlibaba

72.80%

GPT-5.2 CodexOpenAI

72.40%

Grok 4.20 0309 ReasoningxAI

72.20%

Grok 4.3xAI

71.40%

GLM 5 ThinkingZhipu AI

71.40%

Qwen3.5 Plus ThinkingAlibaba

71.20%

Mimo V2.5Xiaomi

71.00%

Kimi K2.5 ThinkingMoonshot AI

70.00%

Claude Sonnet 4.5 20250929 ThinkingAnthropic

70.00%

Qwen3.6 27bAlibaba

70.00%

GPT-5.4 NanoOpenAI

69.80%

GPT-5.1OpenAI

69.80%

GLM 4.7Zhipu AI

69.40%

GPT-5OpenAI

69.00%

Nemotron 3 Ultra 550b A55bNvidia

69.00%

Qwen3.7 MaxAlibaba

68.80%

DeepSeek V3p2 ThinkingFireworks AI

67.60%

Claude Haiku 4.5 20251001 ThinkingAnthropic

66.60%

Mistral Medium 3.5Mistral AI

66.40%

Qwen3.5 FlashAlibaba

64.40%

Gemini 3.1 Flash Lite PreviewGoogle

62.80%

Devstral 2512Mistral AI

62.80%

GPT-5 MiniOpenAI

60.80%

Kimi K2 ThinkingMoonshot AI

60.20%

Grok 4 0709xAI

57.80%

Laguna M.1Poolside

57.60%

Laguna Xs.2Poolside

55.20%

Gemini 2.5 ProGoogle

54.40%

Grok 4 Fast ReasoningxAI

45.40%

Grok 4.1 Fast ReasoningxAI

41.40%

Mistral Large 2512Mistral AI

41.40%

GPT Oss 120bFireworks AI

33.60%

Command A 03 2025Cohere

7.80%

The published Vals SWE-bench mirror snapshot places Claude Opus 5 first at 97.00%. The third row is 2.00 points behind. The broader top-10 range is 14.40 points, so the table still separates the published systems.

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

About Vals SWE-bench mirror

Year

2026

Tasks

Software engineering issue-resolution tasks

Format

Accuracy score

Difficulty

Production software engineering

BenchLM keeps this separate from its canonical SWE-bench Verified page so Vals-hosted results remain secondary context rather than source-of-record data.

Vals SWE-bench Public benchmark source

BenchLM freshness & provenance

Version

Vals SWE-bench mirror 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.

FAQ

What does Vals SWE-bench mirror measure?

Vals AI hosted SWE-bench view for solving production software engineering tasks.

Which model leads the published Vals SWE-bench mirror snapshot?

Claude Opus 5 currently leads the published Vals SWE-bench mirror snapshot with 97.00% vals swe-bench score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on Vals SWE-bench mirror?

75 AI models are included in BenchLM's mirrored Vals SWE-bench mirror snapshot, based on the public leaderboard captured on July 22, 2026.

Last updated: July 22, 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.