Step 3.7 Flash

Composite indexes

• Artificial Analysis Intelligence Index#66 of 142
  20independentv4.3.2 (AA estimate; independent evaluation forthcoming), high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: Claude Opus 5.5 at 58

Coding

• Terminal-Bench 2.x#71 of 103
  59.5%vendorTerminal-Bench 2.1Hugging Face: stepfun-ai/Step-3.7-Flash ↗
  39.3%independentAA, Terminal-Bench 2.1, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: Claude Fable 5.1 at 91.4%

• SciCode#61 of 94
  43.9%independentAA, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: Claude Opus 5.5 at 66.9%

• SWE-bench Pro#18 of 31
  56.3%vendorHugging Face: stepfun-ai/Step-3.7-Flash ↗

  Leader: Claude Opus 5.5 at 89.9%

Agents and tool use

• τ²-Bench#3 of 88
  98.5%independentAA, telecom, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: GLM-5.2 at 99.1%

• GDPval-AA#62 of 103
  845independentGDPval-AA v2.1, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: Claude Opus 5.5 at 1846

Reasoning and knowledge

• Humanity’s Last Exam#71 of 147
  48.1%vendorwith toolsHugging Face: stepfun-ai/Step-3.7-Flash ↗
  21.4%independentAA, no tools, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: Claude Opus 5.5 at 61.4%

• GPQA Diamond#71 of 142
  80.9%independentAA, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: GPT-6 Astra at 96.1%

• CritPt#61 of 139
  2.3%independentAA, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: GPT-5.6 Sol at 32.3%

Vision and long context

• AA Long Context Reasoning#62 of 139
  73.7%independentAA, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: Kimi K3 at 88.7%

• MMMU-Pro#35 of 72
  75.3%independentAA, high effortArtificial Analysis: Step 3.7 Flash ↗

  Leader: Claude Opus 5.5 at 87.7%

Other published results