Industry2 benchmarks

Software Engineering

Shipping working code against real repositories: bug fixes, feature patches, and competitive programming under tests.

The Software Engineeringscore is the mean of a model’s normalized 0–100 scores (direction-aware, so lower-is-better metrics are inverted) across the 2 benchmarks below — the same figure the leaderboard’s industry view ranks by.

Leaders

Claude Opus 4.8 leads this industry with a score of 88.6.

Claude Opus 4.888.6

Claude Opus 4.787.6

DeepSeek V4-Pro87.1

Qwen3.7 Max86.0

DeepSeek V4-Flash85.3

Kimi K2.684.9

Claude Opus 4.680.8

Gemini 3.1 Pro80.6

Qwen3.6-27B80.6

Nemotron 3 Ultra80.5

GPT-5.280.0

Claude Sonnet 4.679.6

DeepSeek V3.278.2

Gemini 3 Flash78.0

Kimi K2 Thinking77.2

Qwen3.6-35B-A3B76.9

Gemini 3 Pro76.2

Claude Haiku 4.573.3

Nova 2 Pro68.1

Gemini 2.5 Pro64.3

Llama 4 Maverick43.4

Llama 4 Scout32.8

Benchmarks in this score

Each model’s scores on these are normalized and averaged to produce the industry score above.

Agentic Coding%

SWE-bench Verified

A human-validated subset of real GitHub issues; the model must produce a patch that passes the repository's tests. Figures are vendor-reported unless noted.

Coding%

LiveCodeBench

Contamination-resistant competitive-programming problems collected over time to avoid training-set overlap.