Claude Sonnet 4.6

ModelActive

by Anthropic · family “claude-sonnet”

Sonnet 4.6 is Anthropic's most capable Sonnet-class model yet, with frontier performance across coding, agents, and professional work. It excels at iterative development, complex codebase navigation, end-to-end project management with... (description from the OpenRouter listing)

in: textin: imagein: fileout: textReasoningTool useStructured output

Epoch Capabilities Index

152.3 #38 of 268

90% CI 149.7 – 154.7

Listing ↗

Released

17 Feb 2026

Context window

1M tokens

Max output

128K tokens

Input $ / 1M tokens

$3.00

Output $ / 1M tokens

$15.0

Cached input $ / 1M

$0.30

Each benchmark shown separately with its own source

Benchmark results (13)

All benchmarks

Benchmark	Domain	Score	vs best recorded	Setting	Run	Source
LMCA	agents	46.5%	71%	max	—	External ↗
DTBench	reasoning	89.9%	91%	max	—	External ↗
OSWorld 2.0	agents	8.3%	26%	max	—	External ↗
DeepSWE	coding	29.9%	40%	high	—	External ↗
ProofBench	math	45.0%	45%	max	—	External ↗
APEX-Agents	agents	43.0%	63%	high	—	External ↗
Chess Puzzles	games	5.0% ±2.2	7%	high	13 Jul 2026	Eval log ↗
SimpleQA Verified	knowledge	35.5% ±1.5	47%	high	10 Aug 2026	Epoch ↗
DeepResearch Bench	agents	54.9%	99%	high	—	External ↗
ARC-AGI-2	reasoning	60.4%	64%	high	—	External ↗
OTIS Mock AIME 2024-2025	math	75.6% ±6.5	76%	high	13 Jul 2026	Eval log ↗
GPQA diamond	science	83.3% ±2.7	87%	high	13 Jul 2026	Eval log ↗
ARC-AGI	reasoning	86.5%	88%	high	—	External ↗

Source: Epoch AI Benchmarking Hub (CC BY 4.0). “External” rows are leaderboard results Epoch collects from third parties. Best reported setting per benchmark is shown.

API list price over time

Price history

$ per 1M tokens

8 observations (OpenRouter listing + Internet Archive snapshots). History accumulates with every ingest run; a single point means no change has been observed yet.