priorsBuilder

Anthropic · released 2024-03-04

Haiku 3

Measured on 20 benchmarks.

General frontier

Not ranked in this mix — too little measured data for its benchmarks.

Your mix’s benchmarks

3 of 8 measured

reasoning

Humanity's Last Exam

587 results
  1. #525none of 5874.1%

Leader: Opus 5.5 · 61.4%

knowledge

AA-Omniscience

526 results
  1. #364none of 526-48.6

Leader: Opus 5.5 · 46.4

reasoning

BullshitBench v2

196 results
  1. #174medium of 19610.0%

Leader: Opus 4.8 · 95.0%

All 20 benchmarks

show 17 more

agentic

τ²-Bench Telecom

393 results
  1. #315none of 39321.1%

Leader: GLM 5.2 · 99.1%

coding

LiveCodeBench

282 results
  1. #233none of 28215.4%

Leader: Gemini 3 Pro · 91.7%

coding

Terminal-Bench Hard-AA

387 results
  1. #334none of 3870.8%

Leader: GPT-5.6 Sol · 65.9%

coding

WeirdML v2

127 results
  1. #124medium of 1279.8%

Leader: GPT-6 Astra Pro · 93.6%

composite

AA Intelligence Index v4.3.2

610 results
  1. #543none of 6105.6

Leader: Opus 5.5 · 57.6

composite

Arena Text

220 results
  1. #199medium of 2201262

Leader: Gemini 4 · 1525

composite

Arena Text — Multi-Turn

220 results
  1. #198medium of 2201245

Leader: Gemini 4 · 1553

instruction

IFBench

399 results
  1. #295none of 39936.1%

Leader: Grok 4.3 · 83.3%

knowledge

AA-Omniscience Accuracy

526 results
  1. #344none of 52617.6%

Leader: Fable 5.1 · 67.2%

knowledge

AA-Omniscience Non-hallucination

526 results
  1. #279none of 52619.5%

Leader: MiniCPM5-1B (Non-reasoning) · 99.1%

knowledge

Global-MMLU-Lite

118 results
  1. #109none of 11863.3%

Leader: Gemini 3.1 Pro · 93.2%

knowledge

GPQA Diamond

539 results
  1. #468none of 53937.4%

Leader: GPT-6 Astra · 96.3%

knowledge

MMMU-Pro

265 results
  1. #257none of 26530.8%

Leader: Opus 5.5 · 87.7%

long-context

AA-LCR v1.1

523 results
  1. #385none of 52327.7%

Leader: Kimi K3 · 88.7%

math

MATH-500

162 results
  1. #147none of 16239.4%

Leader: GPT-5 · 99.4%

other

Arena Vision

116 results
  1. #114medium of 1161000

Leader: Fable 5 · 1308

reasoning

CritPt

529 results
  1. #484none of 5290.0%

Leader: Muse Wonder Full (Max) · 32.9%