Claude Sonnet 5.5: same $2/$10 as Sonnet 5, Anthropic says 30%+ faster

Anthropic released Claude Sonnet 5.5 (blog 2026-09-28), second model in the Claude 5.5 family after Opus 5.5. Primary model ID on the Claude Platform: claude-sonnet-5-5. Anthropic positions it as a faster, lower-cost complement to Opus for well-scoped everyday work (Anthropic).

This is a Desk Bot models briefing.

Pricing (HARD)

Same per-token list as Sonnet 5: $2 / 1M input, $10 / 1M output, $0.20 / 1M cache reads. Anthropic also lists cache writes at $2.50 / 1M (5-minute) on the Sonnet 5.5 pricing table (Anthropic, claude.dev).

Soft: speed, cost-per-task, benches (attribute)

All of the following are Anthropic-stated / self-reported—not independent validation (Anthropic):

• Generates outputs 30%+ faster than Sonnet 5
• Up to 30% less cost per task vs predecessor (fewer tokens for the same work at unchanged list prices)
• Terminal-Bench 4.0: 70.6%; CursorBench 4.0: 55.5% (Anthropic table)

Availability + Haiku teaser

Available on Amazon Web Services, Google Cloud, and Microsoft Azure, plus Claude Platform / Claude Code. Bedrock soft ID (builder hub): anthropic.claude-sonnet-5-5 (Anthropic, claude.dev).

Claude Haiku 5.5 — “coming weeks”; no ship date (Anthropic).

Builder hub sidebar (not a separate post)

From claude.dev only: effort defaults (high on Claude Platform / API; medium in Claude Code); 1M context native; 128k max output; knowledge cutoff June 2026.

Safeguards (high-level)

Anthropic: first Sonnet with cyber safeguards comparable to its most capable models; biology safeguards same as Sonnet 5. Targets a narrow set of high-risk requests; routine software development and most life-sciences work described as unaffected—no PoC/exploit detail here (Anthropic).

Who should care

Teams on Sonnet 5 who want the same list prices with Anthropic-claimed speed/efficiency gains should start at anthropic.com/claude-sonnet-5-5—lock claude-sonnet-5-5, soft-attribute benches, and treat Haiku 5.5 as coming-weeks only.