Skip to content
Company desk / checked September 30, 2026

Anthropic: what changed, and what it means for you

Claude models, plans and Claude Code. What changed, what it costs you, and what to do.

This week in one paragraph

Anthropic shipped two strong models in a week—Opus 5.5 and Sonnet 5.5—at unchanged or lower prices, and made Opus the default on Pro in Claude Code. Fable stays the premium model, pay-as-you-go on Pro.

+ 3 good for readers − 1 costs or risks

The ledger · newest first

Good news and bad news, dated

  1. + Plus

    Sonnet 5.5: faster and cheaper per task, same price

    $2 / $10 per million tokens, 1M context. Anthropic reports scores close to Opus 5.5 on several benchmarks and up to 30% lower cost per task than Sonnet 5.

    For you: Startups: a strong model you can afford to put in your product.

  2. Watch

    Sonnet 5.5 falls back to Sonnet 5 on high-risk cyber requests

    The first Sonnet with Opus-style cyber safeguards. Routine bug fixing is unaffected.

    For you: Security teams: expect visible model switches on some requests.

  3. + Plus

    Opus 5.5 launches; Claude Code makes it Pro’s default

    Claude Code 2.1.280 made Opus 5.5 the default Opus and switched Pro and Team Standard from Sonnet to Opus by default.

    For you: Pro users get Anthropic’s daily-driver model without paying more.

  4. + Plus

    Opus 5.5 reaches Kiro’s paid tiers

    Experimental support for Pro, Pro+, Pro Max and Power at a 2.0× credit multiplier.

    For you: Kiro users can try Opus 5.5 without a Claude plan.

  5. − Minus

    Fable on Pro is pay-as-you-go only

    The all-plans Fable 5 promotion ended July 19. Fable 5 and 5.1 are included on Max and premium seats (up to 50% of weekly limits); Pro and Team Standard use usage credits.

    For you: On $20 Pro, start with Opus 5.5 or Sonnet 5.5.

  6. Watch

    The next Fable is rumored, not announced

    Leakers point to “Fable 5.2”; some say “5.5”. No date or name from Anthropic.

    For you: Don’t buy credits in anticipation of a model that hasn’t shipped.

Models now · API list prices per 1M tokens (input / output)

Which Anthropic model for which job

Claude Fable 5.1$10 / $50Demanding reasoning, long-horizon agentsPrice checked Sep 29
Claude Opus 5.5$4 / $20Daily driver for agentic coding and knowledge workPrice checked Sep 29Claude Sonnet 5.5$2 / $10Speed and cost; well-scoped everyday workPrice checked Sep 29
Claude Haiku 4.5$1 / $5Fastest; Haiku 5.5 “in the coming weeks”Price checked Sep 29
Our read · opinion

What to do now

Sonnet 5.5 makes a frontier-class model affordable in your product; price your features around task cost, not token price.

The OpenAI desk →Personal agents compared →The China Desk →
Today: OpenAI dots