Skip to content
aiqbench

Google DeepMind · model passport

Google: Gemini 2.5 Pro

gemini-2-5-pro
  • Cloud only
  • Proprietary licence
  • 1,048,576 context
  • Released 2025-06-05(curated)
18/27axes covered
Overall · LMArena
1,445#77 of 396
Best percentile
95.1MMLU-Pro (Artificial Analysis)
Composite standing
56.8shrunk · 18/27 axes

What the data says

Benchmarked on 18 of 27 tracked axes. Every figure below is the model's own published result on that board.

Strongest axes

  • 95.1 pct — MMLU-Pro (Artificial Analysis) (86.2%)
  • 90.5 pct — LiveCodeBench (Artificial Analysis) (80.1%)

Weakest axes

  • 7.9 pct — WebDev Arena (1,226)
  • 9.4 pct — SWE-bench Verified (57.6%)
  • 13.2 pct — Document Arena (1,431)

Uneven: 87.2 percentile points between its best and worst axis — 95.1 pct on MMLU-Pro (Artificial Analysis) versus 7.9 pct on WebDev Arena.

  • MMLU-Pro (Artificial Analysis)95.1 pct86.2%Verify →
  • LiveCodeBench (Artificial Analysis)90.5 pct80.1%Verify →
  • Instruction following81.8 pct1,438Verify →
  • Overall80.8 pct1,445Verify →
  • Factuality Arena77.9 pct1,462Verify →
  • Hard prompts77 pct1,458Verify →
  • HHEM Hallucination (Vectara)70.5 pct93.0%Verify →
  • Coding69.8 pct1,464Verify →
  • GPQA Diamond68.8 pct85.3%Verify →
  • Vision Arena68 pct1,248Verify →
  • AIME (math)60.5 pct84.2%Verify →
  • IFBench (Artificial Analysis)55.2 pct48.7%Verify →
  • FrontierMath50.7 pct14.1%Verify →
  • ARC-AGI-234.8 pct4.9%Verify →
  • Terminal-Bench31.8 pct32.6%Verify →
  • Document Arena13.2 pct1,431Verify →
  • SWE-bench Verified9.4 pct57.6%Verify →
  • WebDev Arena7.9 pct1,226Verify →
  • Search ArenaNot benchmarked
  • Agent ArenaNot benchmarked
  • SimpleQA VerifiedNot benchmarked
  • Humanity's Last ExamNot benchmarked
  • SimpleBenchNot benchmarked
  • LiveBenchNot benchmarked
  • BFCL v4 (function calling)Not benchmarked
  • SWE-bench Verified (live board)Not benchmarked
  • Terminal-Bench 4.0Not 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

Use it hosted

API model IDgoogle/gemini-2.5-pro
Context length1,048,576 tokens
Max output65,536 tokens
Acceptstext, image, file, audio, video
Input price$1.25 per 1M tokens
Cached input price$0.13 per 1M tokens
Output price$10 per 1M tokens

API parameters this endpoint supports

Include reasoningMax tokensReasoningResponse formatSeedStop sequencesStructured outputsTemperatureTool choiceTool callingTop-p

Example request

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"google/gemini-2.5-pro","messages":[{"role":"user","content":"Hello"}]}'

View on OpenRouter →

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.

  • gemini-2.5-pro stands at 56.8 (raw mean 58) — measured on 18 of 27 axes.

    standingtemplateVerify →

Every claim is a computed fact from the current board data, phrased by a low-temperature model call and rejected by a validation gate if any number is not in the source row. How the leaderboards work →