← leaderboard

Claude Haiku 4.5

CALIBRATING

Anthropic · anthropic/claude-haiku-4.5 · verdict as of 2026-07-03

Daily test history (90 days, baseline band = trailing mean ± 2σ)

Math & Reasoning

CALIBRATING

Code Generation

CALIBRATING

Game Design

CALIBRATING

Instruction Following

CALIBRATING

Report Analysis

CALIBRATING

Structured Extraction

CALIBRATING

Summarization

CALIBRATING

Customer Service

CALIBRATING

Web Design

CALIBRATING

Creative Writing

CALIBRATING

Serving providers (last 30 days)

striped = multiple providers served this model that day (hover for detail)

Public benchmarks overall 65.2

MMLU-Pro

81

GPQA Diamond

73

SWE-bench Verified

66

LMArena Elo

1380

AIME 2025

75

retrieved 2026-07-02 from public sources — see methodology

Recent samples (latest run, one per test case)