Claude Sonnet 4.6

Claude Sonnet 4.6 is the best value proposition in the Claude lineup — 79.6% on SWE-bench Verified (just 1.2 points behind Opus) at 40% of the cost.

Benchmarks

Benchmark	Score	vs Opus 4.6
SWE-bench Verified	79.6%	-1.2 pts
GPQA Diamond	88.5%	-5.8 pts
LM Arena Elo	1478	-26 pts

The gap between Sonnet and Opus is the narrowest it has ever been. For most tasks, the quality difference is imperceptible.

Pricing

	Per 1M tokens	vs Opus
Input	$3.00	40% cheaper
Output	$15.00	40% cheaper

Architecture & capabilities

• Context: 1M tokens — same as Opus, at standard pricing
• Output: Up to 64K tokens
• Thinking modes: Adaptive, extended, and interleaved thinking — same as Opus
• Shares all core Opus improvements

Strengths

• Best price/performance ratio in the frontier tier
• 1M context at $3/$15 — cheaper than most competitors
• Same thinking capabilities as Opus
• Strong enough for daily coding, analysis, and agentic work

Weaknesses

• Slightly weaker on graduate-level science reasoning (GPQA gap)
• Lower LM Arena ranking — noticeable in head-to-head comparisons

When to use

Default choice for most tasks. Use Opus only when you need the absolute best reasoning quality or are working on PhD-level scientific analysis.