Google: Gemini 3.1 Pro Preview

What the data says

Benchmarked on 20 of 27 tracked axes. Every figure below is the model's own published result on that board.

Uneven: 70.4 percentile points between its best and worst axis — 97.7 pct on SimpleBench versus 27.3 pct on Agent Arena.

• SimpleBench97.7 pct79.6%Verify →
• GPQA Diamond96.8 pct94.4%Verify →
• SimpleQA Verified96.3 pct73.5%Verify →
• Overall96.2 pct1,487Verify →
• LiveBench96.2 pct77.1%Verify →
• Hard prompts96.2 pct1,508Verify →
• Humanity's Last Exam95.7 pct46.4%Verify →
• Instruction following95.7 pct1,480Verify →
• Terminal-Bench95.5 pct80.2%Verify →
• IFBench (Artificial Analysis)95.4 pct77.1%Verify →
• Coding92.8 pct1,521Verify →
• Factuality Arena87.2 pct1,472Verify →
• Vision Arena84.7 pct1,280Verify →
• ARC-AGI-283.1 pct77.1%Verify →
• FrontierMath82.6 pct36.9%Verify →
• AIME (math)82 pct95.6%Verify →
• WebDev Arena54 pct1,447Verify →
• HHEM Hallucination (Vectara)42.9 pct89.6%Verify →
• Document Arena42.1 pct1,459Verify →
• Agent Arena27.3 pct-0.040Verify →
• Search ArenaNot benchmarked
• BFCL v4 (function calling)Not benchmarked
• SWE-bench Verified (live board)Not benchmarked
• Terminal-Bench 4.0Not benchmarked
• MMLU-Pro (Artificial Analysis)Not benchmarked
• LiveCodeBench (Artificial Analysis)Not benchmarked
• SWE-bench VerifiedNot benchmarked

Percentiles are the model's standing within each board's own population. Axes marked "not benchmarked" have no published result for this model — that is an absence of data, not a low score.

How to run it

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"google/gemini-3.1-pro-preview","messages":[{"role":"user","content":"Hello"}]}'

Identifiers, limits and prices are synced from the upstream source shown above. Any field we haven't synced is left out entirely rather than guessed.