New featureFeaturedJuly 1, 2026
Claude Sonnet 5 is live on Napsix — the new default for everyday work
Anthropic's most agentic Sonnet: near Opus 4.8 quality at Sonnet price (with launch pricing). Already routing chat, agents and AI features on paid plans.
aimodelosanthropicsonnet-5
Anthropic released Claude Sonnet 5 (June 30) and it's already integrated across the platform.
Why it matters
It's the most agentic Sonnet yet — the first Sonnet to reach frontier-level quality for most real work:
| Benchmark | Sonnet 5 | Sonnet 4.6 | Opus 4.8 |
|---|---|---|---|
| SWE-bench Pro (agentic coding) | 63.2% | 58.1% | 69.2% |
| Terminal-Bench 2.1 (terminal + tools) | 80.4% | 67.0% | 82.7% |
| OSWorld (computer use) | 81.2% | 78.5% | 83.4% |
| GDPval (knowledge work) | 1618 Elo | 1395 | 1615 |
On knowledge work it beats Opus 4.8, and stays within a few points everywhere else — at a fraction of the cost.
What changes on Napsix
- New default for paid plans: XIA chat, Rooms/Projects agents, Mission Control and AI features (summaries, translation, OCR, code) that used Sonnet 4.6 now route to Sonnet 5 automatically. Nothing for you to change.
- 1M context window (5× Sonnet 4.6) and 128K output.
- Launch pricing: ~33% cheaper than Sonnet 4.6 until August 31, then same price.
- Sonnet 4.6 stays available as the automatic fallback.
As always: you don't pick the model. Ask to "write an email with AI" or chat with XIA, and the platform uses the best one. Technical details at anthropic.com/news/claude-sonnet-5.
