Skip to content
aiqbench

Anthropic · model passport

Anthropic: Claude Sonnet 4.5

claude-sonnet-4-5
  • Cloud only
  • Proprietary licence
  • 1,000,000 context
  • Released 2025-09-29(curated)
20/27axes covered
Overall · LMArena
1,455#61 of 396
Best percentile
98.8BFCL v4 (function calling)
Cohort standing
59.7median 52.6

What the data says

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

Strongest axes

  • 98.8 pct — BFCL v4 (function calling) (73.2%)
  • 89.4 pct — Instruction following (1,462)
  • 89.2 pct — SWE-bench Verified (live board) (74.8%)

Uneven: 73.2 percentile points between its best and worst axis — 98.8 pct on BFCL v4 (function calling) versus 25.6 pct on SimpleQA Verified.

  • BFCL v4 (function calling)Dormant board98.8 pct73.2%Verify →
  • Instruction following89.4 pct1,462Verify →
  • SWE-bench Verified (live board)89.2 pct74.8%Verify →
  • Coding87.7 pct1,512Verify →
  • LiveBench87.3 pct72.3%Verify →
  • Hard prompts86.9 pct1,483Verify →
  • Overall84.8 pct1,455Verify →
  • Factuality Arena64.5 pct1,449Verify →
  • GPQA Diamond63.3 pct82.3%Verify →
  • Terminal-Bench59.1 pct46.5%Verify →
  • SimpleBench58.6 pct54.3%Verify →
  • AIME (math)56 pct77.8%Verify →
  • FrontierMath52.2 pct15.2%Verify →
  • Humanity's Last Exam47.8 pct13.7%Verify →
  • ARC-AGI-247.2 pct13.6%Verify →
  • WebDev Arena36.7 pct1,394Verify →
  • SWE-bench Verified31.3 pct71.3%Verify →
  • Document Arena28.9 pct1,449Verify →
  • HHEM Hallucination (Vectara)27.6 pct88.0%Verify →
  • SimpleQA Verified25.6 pct30.7%Verify →
  • Vision ArenaNot benchmarked
  • Search ArenaNot benchmarked
  • Agent ArenaNot benchmarked
  • Terminal-Bench 4.0Not benchmarked
  • MMLU-Pro (Artificial Analysis)Not benchmarked
  • LiveCodeBench (Artificial Analysis)Not benchmarked
  • IFBench (Artificial Analysis)Not 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 IDanthropic/claude-sonnet-4.5
Context length1,000,000 tokens
Max output64,000 tokens
Acceptstext, image, file
Input price$3 per 1M tokens
Cached input price$0.30 per 1M tokens
Output price$15 per 1M tokens

API parameters this endpoint supports

Include reasoningMax completion tokensMax tokensReasoningResponse formatStop sequencesStructured outputsTemperatureTool choiceTool callingTop-kTop-p

Example request

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"anthropic/claude-sonnet-4.5","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.

  • claude-sonnet-4-5-20250929 stands at 59.7 (raw mean 61.2) — measured on 20 of 27 axes.

    standingtemplateVerify →

  • claude-sonnet-4-5-20250929 stands at 59.7 versus a cohort median of 52.6 across 143 models released within 180 days — above the peer median.

    cohorttemplateVerify →

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 →