— reading now
Flash

TopicClaude 5.5 models

Story·Models & Products·2026-09-29 03:30

Anthropic Releases Claude Sonnet 5.5: Same Sticker Price, ~30% Cheaper Per Task

Anthropic today released Claude Sonnet 5.5 — the second model in its 5.5 family. Same sticker price as Sonnet 5: $2 per million input tokens, $10 per million output tokens. But Anthropic says the same work now takes far fewer tokens, with output over 30% faster. The company is preparing for an IPO.

What Happened

On Monday, September 28, Reuters reported that Anthropic officially released Claude Sonnet 5.5, following last week's Opus 5.5. The company calls it the fastest Sonnet to date, positioned as the workhorse for everyday enterprise tasks: coding, document creation, spreadsheets.

Sticker price holds at $2/$10, but Anthropic says real token consumption per task drops sharply — combined with 30%+ faster output, per-task cost falls by roughly a third. Early testers gave numbers too: Box reported 2.4x speed with 12% fewer tokens; Atlassian said Rovo Agents ran about 30% faster.

Opus 5.5 ($4/$20) last week, Sonnet 5.5 this week, and Haiku 5.5 "coming soon" — all three tiers refreshed within weeks, as the company prepares for an IPO. Enterprise customers now account for about 80% of Anthropic's business, with Salesforce, Databricks, Goldman Sachs and Novo Nordisk on the roster.

Key Facts

  1. Release: Claude Sonnet 5.5, second in the 5.5 family; billed as the fastest Sonnet ever, with far fewer tokens per task.
  2. Pricing: Unchanged: $2 input / $10 output per 1M tokens; roughly 30% cheaper per task.
  3. Measured: Box: 2.4x speed, 12% fewer tokens; Atlassian: Rovo Agents ~30% faster. (Source: Unite.AI)
  4. Positioning: Everyday enterprise work: coding, docs, spreadsheets; first Sonnet with cyber-grade safeguards and fallbacks.
  5. Cadence: Opus 5.5 last week → Sonnet 5.5 this week → Haiku 5.5 due soon; the company is gearing up for IPO.

Context

OpenAI just launched cheaper GPT-6 Sol/Luna models, halving API prices; meanwhile more enterprises are switching to open-weight models to cut costs — the Financial Times cited AlphaSense data showing executive mentions of open models jumped sixfold year-over-year in August–September.

OpenAI cut prices directly; Anthropic kept its sticker price and lowered the cost per task.

Anthropic CEO Dario Amodei publicly called this month to "slow down the pace" of capability releases over safety concerns; Anthropic has refreshed all three model tiers within weeks.

Why it matters

The $2/$10 price is unchanged, but tasks use fewer tokens and output is 30%+ faster, cutting per-task cost by about a third; Box measured 2.4x speed with 12% fewer tokens.
Useful Tap if this story helped you

SourcesReuters (Sept 28) · 9to5Mac · VentureBeat · Unite.AI · Financial Times (via CoinCentral). Compiled from public reporting; not investment advice.

Comments

  1. Loading comments…
Ask the cat