modelbenchmark.io

Claude Haiku 4.5

anthropic-claude-4-5-haiku

Compare

Fast Claude model for responsive assistance, classification, and lightweight agents

Specification

most-agreed values

Context
200K
Max output
64K
Released
2025-10-15
Knowledge cutoff
2025-02-28
Retires
Open weights
no
Input
text, image
Output
text

Price

US dollars per million tokens · most-agreed

Input
$1.00
Output
$5.00
Cache read
$1.00
Cache write
$1.25

Quality

7 benchmarks · 1 source

Composite
44th
percentile of 334 scored models
Rank
186 / 334
±0.06 sd
Evidence
7 × 1
one source, or fewer than 3 benchmarks
Effort range
9.5
points between effort settings
reasoning50th2/4 bench
math55th4/7 bench
knowledge10th1/1 bench

Every score

one row per source and configuration — nothing averaged away

BenchmarkScoreConfigurationSourceRun
MATH level 596.4 ±0.532kEpoch AI2025-10-22
86.9 ±0.8defaultEpoch AI2025-10-16
Settings differ by 9.5 points on this benchmark.
GPQA diamond71.2 ±3.232kEpoch AI2025-10-22
60.5 ±2.8defaultEpoch AI2025-10-16
Settings differ by 10.7 points on this benchmark.
OTIS Mock AIME 2024-202566.7 ±7.132kEpoch AI2025-10-22
35.8 ±6.2defaultEpoch AI2025-10-16
Settings differ by 30.9 points on this benchmark.
FrontierMath-Tier-4-2025-07-012.1 ±2.132kEpoch AI2025-10-22
Chess Puzzles8.0 ±2.732kEpoch AI2026-07-16
FrontierMath-v14.1 ±1.2defaultEpoch AI2025-10-16
3.0 ±1.432k · 2 runsEpoch AI2025-10-22
Settings differ by 1.1 points on this benchmark.
SimpleQA Verified13.2 ±1.1defaultEpoch AI2026-08-10
12.6 ±1.132kEpoch AI2026-08-27
Settings differ by 0.6 points on this benchmark.

The leading figure for a benchmark is the median across its configurations, so one heroic high-effort run cannot set the number.

Available from 2 hosts

HostIn $/MOut $/MCache rdCache wrContextOutputRetires
digitalocean$1.00$5.00$1.00$1.25200K64K
sap-ai-core$1.00$5.00$0.10$1.25200K64K