SWE-bench

Indexed results from SWE-bench. Each ranking retains its own metric and evaluation conditions.

SWE-bench Lite

sweBench · SWE-bench · Resolved tasks · percent

  1. OpenAI: GPT-4o (2024-08-06)openai36%
  2. OpenAI: GPT-4o (2024-08-06)openai21.67%

SWE-bench Multilingual

sweBench · SWE-bench · Resolved tasks · percent

  1. claude-opus-4-5@20251101anthropic70.7%
  2. OpenAI: GPT-5.2openai66.7%
  3. claude-haiku-4-5@20251001anthropic64.7%
  4. OpenAI: GPT-5 Miniopenai39.7%

SWE-bench Multimodal

sweBench · SWE-bench · Resolved tasks · percent

  1. OpenAI: o3openai35.98%
  2. Claude Sonnet 4anthropic34.43%
  3. OpenAI: o4 Mini (batch)openai33.85%
  4. OpenAI: GPT-4.1openai31.14%
  5. OpenAI: GPT-4oopenai11.99%
  6. OpenAI: GPT-4oopenai9.28%

SWE-bench Verified

sweBench · SWE-bench · Resolved tasks · percent

  1. claude-opus-4-5@20251101anthropic74.4%
  2. gemini-3-pro-previewgoogle74.2%
  3. OpenAI: GPT-4oopenai71.2%
  4. OpenAI: GPT-5openai71.2%
  5. Anthropic: Claude Sonnet 4.5anthropic70.6%
  6. gemini-3-pro-previewgoogle69.6%
  7. OpenAI: GPT-5.2openai69%
  8. claude-haiku-4-5@20251001anthropic66.6%
  9. OpenAI: GPT-5.1openai66%
  10. MoonshotAI: Kimi K2 0711moonshotai65.4%
  11. OpenAI: GPT-5openai65%
  12. OpenAI: o3openai58.4%

Models

Loading
EVIDENCE / COMPARISON

Benchmarks

Explore every benchmark

Loading

These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

Install TensorTrace