Claude Sonnet 4.6

Claude Sonnet 4.6 is Anthropic's most capable Sonnet model, bringing Opus-level intelligence to the balanced Sonnet tier with major improvements in coding, computer use, long-context reasoning, and design.

Claude Sonnet 4.6 is our most capable Sonnet model yet. It's a full upgrade of the model's skills across coding, computer use, long-context reasoning, agent planning, knowledge work, and design.

Anthropic

Overview

Released February 17, 2026, Sonnet 4.6 closes the gap with Opus-class models while maintaining Sonnet's practical price point. In early testing, developers preferred Sonnet 4.6 over Sonnet 4.5 roughly 70% of the time—and even preferred it to Opus 4.5 59% of the time, citing better instruction following, less overengineering, and more consistent follow-through on multi-step tasks.

Key Features

1M token context window (beta)
Opus-level performance at Sonnet pricing
Major computer use improvements: Human-level capability on complex spreadsheets, multi-step web forms
Enhanced prompt injection resistance vs Sonnet 4.5
Adaptive and extended thinking support
Context compaction (beta): Automatic summarization as conversations approach limits

Benchmark Highlights

Sonnet 4.6 approaches Opus-level scores across coding, reasoning, and agent benchmarks while showing particular strength in:

Computer use: Steady OSWorld gains, approaching human-level on real software tasks
Long-horizon planning: Novel strategies in Vending-Bench Arena
Document comprehension: Matches Opus 4.6 on OfficeQA
Frontend design: Notably more polished layouts, animations, and design sensibility

When to Use Claude Sonnet 4.6

Choose Sonnet 4.6 when you need:

Best balance of intelligence and cost
Strong coding with less overengineering than Opus
Computer use for real-world automation
Long-context reasoning (up to 1M tokens)
Production workloads at scale
Frontend development with design polish
Document analysis and knowledge work

Choose Opus 4.6 when you need:

Deepest reasoning for the hardest problems
Codebase refactoring at scale
Multi-agent workflow coordination
Tasks where getting it exactly right is paramount

Choose Haiku 4.5 when you need:

Maximum speed and lowest latency
Cost optimization at high volume
Simpler tasks that don't require deep reasoning

What's Improved Over Sonnet 4.5

Based on developer feedback:

Better context reading: More effectively reads context before modifying code
Logic consolidation: Consolidates shared logic rather than duplicating
Instruction following: Significantly improved adherence to directions
Fewer false claims: Less likely to claim success incorrectly
Fewer hallucinations: More grounded responses
Multi-step consistency: Better follow-through on complex tasks
Prompt injection resistance: Major improvement in computer use safety

Role in Series

Claude Sonnet evolution:

Sonnet 3.5 (Jun 2024): Established Sonnet as coding leader
Sonnet 3.5 v2 (Oct 2024): First computer use capabilities
Sonnet 4 (May 2025): Claude 4 architecture
Sonnet 4.5 (Sep 2025): Enhanced reasoning
Sonnet 4.6 (Feb 2026): Opus-level intelligence at Sonnet pricing (this model)

Anthropic: Claude Sonnet 4.6

Model Type