Atlas

Models

← All models

Sonnet 4.5

Anthropic · 2025-09-29

Claude Sonnet 4.5 is Anthropic's hybrid-reasoning model for coding, agentic tasks, and computer use. It supports optional extended thinking; at launch, Anthropic reported it as state of the art on SWE-bench Verified and observed it maintaining focus for more than 30 hours on complex, multi-step tasks.

Benchmark scores

BenchmarkScore
A-Fantasia40.3
A-Fantasia (Valid Attempts with Retries)40.3
A-Fantasia Backwards Spelling24.0
A-Fantasia Chess25.0
A-Fantasia Chess (Valid Attempts with Retries)25.0
A-Fantasia Cube (Valid Attempts with Retries)72.0
A-Fantasia Cube Rotation72.0
A-Fantasia Spelling (Valid Attempts with Retries)24.0
AA-Briefcase Elo712
AA-Briefcase v1.1 Elo714
AA-LCR65.7
AA-LCR v1.172.3
AA-Omniscience Index-0.1
Agent Red Teaming k=100 (CAIS)64.6
AIME 2024 (avg@8)90.7
AIME 2024-202588.2
AIME 2025100.0
AIME 2025 (Artificial Analysis avg@10)88.0
AIME 2025 (avg@8)85.7
AIME 2025 (MathArena repeated-run mean)84.2
ALE-Bench1192
AlgoTune1.5
Apex (MathArena)1.6
APEX-SWE31.0
ARC-AGI-1 Semi-Private63.7
ARC-AGI-213.6
Arena Code WebDev — Brand & Marketing1385
Arena Code WebDev — Consumer Product1385
Arena Code WebDev — Content Creation Tools1383
Arena Code WebDev — Data & Analytics1386
Loading Atlas data…