Source: Claude models overview (Anthropic)

Anthropic's model catalogue: current Claude lineup, IDs, context windows, output limits, prices, cutoffs and retirement dates.

Contents
  • What: official API documentation page
  • Publisher: Anthropic
  • Date: accessed 2026-11-14 (page lists retirement commitments to late 2027)
  • Original: Models overview

Summary

The canonical table of Claude models. Four current tiers are presented (Fable, Opus, Sonnet, Haiku) plus a limited-access Mythos line, with per-platform model IDs (Claude API, Bedrock, Google Cloud, Microsoft Foundry, Claude Platform on AWS), and a Models API that reports token limits and capabilities programmatically.

Key claims

  • As of 2026-11-14 the current lineup is Claude Fable 5.2 ($8/$40 per MTok), Opus 5.5 ($4/$20), Sonnet 5.5 ($2/$10) and Haiku 5.5 (from $0.10/$0.50).
  • All four current models have a 1M-token context window, 128K max output (synchronous), and a reliable knowledge cutoff of June 2026.
  • Default recommendation: start with Opus 5.5; use Fable 5.2 for demanding reasoning and long-horizon agentic work or when Opus 5.5 at higher effort falls short.
  • Comparative latency: Fable and Opus moderate, Sonnet fast, Haiku fastest.
  • Thinking is adaptive (model decides how much to think, steered by an effort parameter); on Fable 5.2 and Opus 5.5 it is always on. Default effort is high on Fable 5.2 and Sonnet 5.5, medium on Opus 5.5 and Haiku 5.5.
  • Batch API is 50% off; cache reads cost 10% of base input (2.5% on Fable 5.x, 5% on Opus 5.5 and Sonnet 5.5).
  • Claude Mythos 5.2 has Fable 5.2’s capabilities but is limited to organisations verified through Anthropic programmes (e.g. cyber verification).
  • Every model ID is a pinned snapshot, including the dateless IDs used from the 4.6 generation on.
  • On the Batch API several current models support up to 300K output tokens with a beta header.
  • Older models remain listed (e.g. Opus 4.5 and Haiku 4.5 at 200K context); some 4.x models are retired on the first-party API but still served on partner clouds.
  • The Models API returns max_input_tokens, max_tokens and a capabilities object, including whether thinking can be disabled.

Feeds

Frontier model landscape · Model selection

Suggest changes →