Gemma 4 12B

Composite indexes

• Artificial Analysis Intelligence Index#78 of 142
  14independentv4.3.2, reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗
  9independentv4.3.2, non-reasoningArtificial Analysis: Gemma 4 12B (Non-reasoning) ↗

  Leader: Claude Opus 5.5 at 58

Coding

• Terminal-Bench 2.x#77 of 103
  27.3%independentAA run (Terminal-Bench 2.1), reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗

  Leader: Claude Fable 5.1 at 91.4%

• LiveCodeBench#22 of 56
  72%vendorLiveCodeBench v6Gemma 4 model card ↗

  Leader: Fugu at 92.9%

Agents and tool use

• GDPval-AA#81 of 103
  413independentGDPval-AA v2.1, reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗

  Leader: Claude Opus 5.5 at 1846

• τ²-Bench#54 of 88
  69%vendoraverage over 3 domainsGemma 4 model card ↗
  36.3%independentAA run, telecom, reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗

  Leader: GLM-5.2 at 99.1%

Reasoning and knowledge

• GPQA Diamond#88 of 142
  78.8%vendorGemma 4 model card ↗
  75.3%independentAA run, reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗

  Leader: GPT-6 Astra at 96.1%

• Humanity’s Last Exam#80 of 147
  15.7%independentAA run, no tools, reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗
  5.2%vendorno toolsGemma 4 model card ↗

  Leader: Claude Opus 5.5 at 61.4%

• CritPt#96 of 139
  0%independentAA run, reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗

  Leader: GPT-5.6 Sol at 32.3%

• MMLU-Pro#10 of 14
  77.2%vendorinstruction-tunedGemma 4 model card ↗

  Leader: Nemotron 3 Ultra 550B A55B at 86.8%

• AIME 2026#10 of 13
  77.5%vendorno toolsGemma 4 model card ↗

  Leader: ERNIE 5.1 at 99.6%

Vision and long context

• AA Long Context Reasoning#82 of 139
  63.7%independentAA-LCR v1.1, reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗

  Leader: Kimi K3 at 88.7%

• MMMU-Pro#44 of 72
  69.7%independentAA run, reasoningArtificial Analysis: Gemma 4 12B (Reasoning) ↗
  69.1%vendorGemma 4 model card ↗

  Leader: Claude Opus 5.5 at 87.7%

Other published results