Claude Haiku 4.5 (20251001) (Thinking): benchmark results

Claude Haiku 4.5 (20251001) evaluated with thinking enabled. Provider: Anthropic. Released 2025-10-01. Access: API.

Unified ELO 1542 ± 1, rank #663 of 1761 rated models, from 38 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
UGI - Writing49.08Writing Score87.9
UGI - Natural Intelligence36.76NatInt Score85.1
Vals AI MedScribe85.23Accuracy (%)76.8
Vals AI MGSM92.15Accuracy (%)72.7
Vals AI AIME82.71Accuracy (%)54.7
Vals AI CorpFin v260.61Accuracy (%)49.6
Vals AI LegalBench81.24Accuracy (%)46.8
Vals AI Terminal-Bench 2.038.2Accuracy (%)46.2
Vals AI CaseLaw v256.48Accuracy (%)44.1
Vals AI MortgageTax62.16Accuracy (%)43.3
Vals AI Finance Agent46.93Accuracy (%)40
Vals AI GPQA72.22Accuracy (%)34.7

Interactive version: theaggregate.ai/model?slug=claude-haiku-4-5-20251001-thinking · How It Works · Data refreshed daily, snapshot 2026-09-05.