Anthropic

Claude 3.5 Haiku benchmark results

7 results on 7 benchmarks from 3 sources.

Closed Released: Oct 22, 2024 Updated Oct 22, 2026 API prices & specs →

Also listed as: claude-3-5-haiku-20241022 · Claude 3.5 Haiku (Oct 2024) · Claude 3.5 Haiku (20241022) · Claude-3-5-Haiku-20241022

All benchmark results

Benchmarks	Score	Rank	Setting	Source type	Date
LMArena Text	1325	250	Overall, style control on; 69064 votes; 95% CI 1321–1328“claude-3-5-haiku-20241022”	Official leaderboard	Oct 8, 2026
LMArena Vision	1127	121	Overall, style control on; 1400 votes; 95% CI 1111–1143“claude-3-5-haiku-20241022”	Official leaderboard	Oct 2, 2026
GPQA Diamond	38.1%	242	Epoch AI run, accuracy; effort: Oct 2024	Independent run	Mar 12, 2025
MMLU	74.3%	41	Exact match; 5-shot; effort: Oct 2024	Official leaderboard via Epoch AI	Oct 22, 2024
MMLU-Pro	62.1%	132	Overall accuracy; data source: TIGER-Lab“Claude-3-5-Haiku-20241022”	Official leaderboard	Mar 11, 2026
AIME (OTIS Mock AIME 2024–2025)	4.3%	237	Epoch AI run, OTIS Mock AIME 2024–2025; effort: Oct 2024	Independent run	Feb 25, 2025
MATH Level 5	46.4%	65	Epoch AI run, accuracy; effort: Oct 2024	Independent run	Mar 12, 2025

← AI Model Benchmarks

FAQ

Where do these Claude 3.5 Haiku scores come from?

From Epoch AI Benchmarking Hub, LMArena leaderboard dataset, and MMLU-Pro leaderboard (TIGER-Lab). Each row links to the original leaderboard, dataset or report, and vendor-reported numbers are labelled.