The new baseline for agentic AI
On June 30, 2026, Anthropic launched Claude Sonnet 5 — simultaneously with Fable 5 return after export controls. The timing is not coincidental: while Fable 5 serves the top of the pyramid, Sonnet 5 redefines what is possible in the middle.
Sonnet 5 immediately became the default model for all Free and Pro plans. API launch price: $2/M input tokens and $10/M output through August 31, 2026 — then rising to $3/$15. Compare to Opus 4.8 ($5/$25) and Fable 5 ($10/$50). Sonnet 5 delivers near-Opus 4.8 agentic performance at one-third the Opus cost and one-fifth the Fable 5 cost.
What changed in practice
Agentic coding: 63.2% benchmark vs Opus 4.8 69.2% and Sonnet 4.6 58.1%. Knowledge work: Sonnet 5 slightly outperforms Opus 4.8. Self-verification: testers report it "finishes complex tasks where previous Sonnet versions stopped" and "checks its own output without being explicitly asked." Real partner example: a team gave Sonnet 5 a two-part job — update Salesforce account tiers, send launch announcement to enterprise contacts — and it finished end-to-end. "It used to stall halfway."
What Sonnet 5 does not replace
Opus 4.8 remains superior for the highest-precision tasks. Important caveat: the new Sonnet 5 tokenizer generates 1.0-1.35x more tokens for the same input — meaning after the introductory price ends in September, Sonnet 4.6 workloads will cost ~30% more when migrated to Sonnet 5.
The strategic context
Sonnet 5 is the first significant model released after Anthropic confidentially filed an S-1 with the SEC on June 1, 2026, following a $65 billion Series H round valuing the company at $965 billion. Democratizing agentic capability to the free tier is also a race for market share before a potential IPO.

