Claude 3.7 Sonnet

Name: Claude 3.7 Sonnet
Author: Anthropic

Anthropic · 2025-02-24

Claude 3.7 Sonnet is Anthropic's first hybrid-reasoning Sonnet model, offering near-instant responses plus an optional extended-thinking mode for harder problems. It targets coding, multimodal tasks, and agentic work.

Benchmark scores

Benchmark	Score
A-Fantasia	45.3
A-Fantasia (Valid Attempts with Retries)	45.3
A-Fantasia Backwards Spelling	35.0
A-Fantasia Chess	33.0
A-Fantasia Chess (Valid Attempts with Retries)	33.0
A-Fantasia Cube (Valid Attempts with Retries)	68.0
A-Fantasia Cube Rotation	68.0
A-Fantasia Spelling (Valid Attempts with Retries)	35.0
AA-LCR	60.7
AA-LCR v1.1	51.7
AA-Omniscience Index	-0.7
Agentic Tool Use (Chat)	56.3
Agentic Tool Use (Enterprise)	65.3
AidanBench	2170
AidanBench (historical 62-prompt subset)	1907
Aider Polyglot	64.9
AIME 2024 (avg@8)	49.2
AIME 2024-2025	44.6
AIME 2025	49.5
AIME 2025 (Artificial Analysis avg@10)	56.3
AIME 2025 (avg@8)	40.0
AIME 2025 (MathArena repeated-run mean)	49.2
ARC-AGI-1 Semi-Private	28.6
ARC-AGI-2	0.9
Arena Text — Business, Management & Financial Ops (No Style Control)	1273
Arena Text — Chinese (No Style Control)	1297
Arena Text — Creative Writing (No Style Control)	1316
Arena Text — English (No Style Control)	1310
Arena Text — Entertainment, Sports & Media (No Style Control)	1300
Arena Text — Exclude Ties (No Style Control)	1241