LLM

Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort)

Name: Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort) Review & Benchmarks
Item: Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort)
Rating: 51.7
Author: Sentient Weekly

by Anthropic

Overall Score

51.7

Anthropic's balanced workhorse — fast, capable, cost-effective

Released Feb 2026

Benchmark Scores

Reasoning

9.0

Coding

50.9

Math

—

Creative writing

8.7

Instruction following

9.2

Multimodal

8.0

Standard Benchmarks

aa_gpqaaa_gpqa

0.9

Measured 2026-05-13 · source

aa_hleaa_hle

0.3

Measured 2026-05-13 · source

aa_scicodeaa_scicode

0.5

Measured 2026-05-13 · source

aa_ifbenchaa_ifbench

0.6

Measured 2026-05-13 · source

aa_lcraa_lcr

0.7

Measured 2026-05-13 · source

aa_terminalbench_hardaa_terminalbench_hard

0.5

Measured 2026-05-13 · source

aa_tau2aa_tau2

0.8

Measured 2026-05-13 · source

Price

$3 / 1M input · $15 / 1M output

Paid

Context

500K tokens

Speed

68.47 t/s

55175ms TTFT

Compare this model →

Overview

The middle tier of the Claude family. Nearly Opus-level capability at a fraction of the cost. Default choice for production workloads where latency and budget matter alongside quality.

Strengths

Excellent quality-to-cost ratio · fast inference · strong coding and reasoning · same training rigor as Opus

Known limitations

Lower ceiling on hardest reasoning tasks vs. Opus · smaller team behind extended thinking improvements

Discussion

0 comments

Be the first to comment.