Claude Sonnet 4.6 Hits Opus-Level Intelligence at Half the Price

Anthropic just dropped Claude Sonnet 4.6, and it's a watershed moment for AI economics. The company's mid-tier model now matches the performance of Claude Opus 4.5--their flagship model from just three months ago--while maintaining Sonnet's significantly lower price point of $3 per million input tokens.

For context: that's the kind of capability jump that used to take 12-18 months, compressed into a quarter. And it fundamentally changes the cost-performance calculation for anyone running AI at scale.

What's Actually Different

Sonnet 4.6 isn't just incrementally better. It's a full rebuild across the model's core competencies:

Computer use jumped from 22.3% accuracy on OSWorld benchmarks (Sonnet 4.5) to 41.6%. That's not a typo. The model can now navigate spreadsheets, fill out multi-step web forms, and coordinate actions across browser tabs with near-human reliability.

Coding performance improved so dramatically that early-access developers prefer Sonnet 4.6 to Opus 4.5 roughly 59% of the time. Users report fewer hallucinations, better instruction following, and less laziness--the tendency of models to cut corners on complex tasks.

What This Means For Your Business

The performance-to-cost ratio here is extraordinary. Work that previously required reaching for an Opus-class model--at $15 per million input tokens--now runs on Sonnet at $3. That's an 80% cost reduction with no performance penalty.

The companies winning with AI right now aren't the ones with the most sophisticated prompts. They're the ones with processes to continuously re-evaluate what's possible and economic.