# Claude Fable 5.2

> Claude Fable 5.2 was released by Anthropic on November 9, 2026, succeeding Claude Fable 5.1 alongside Claude Mythos 5.2, which offers the same specifications and pricing only to verified organizations (such as the Cyber Verification Program). Anthropic builds Fable 5.2 for demanding reasoning and long-horizon agentic work—agent sessions that run for hours, multistep deep research, and analysis carried through to a finished document, spreadsheet, or deck. Its comparative latency improves from Slower on Fable 5.1 to Moderate, and retirement is not sooner than November 9, 2027. For most workloads, Anthropic recommends starting on Claude Opus 5.5 and moving to Fable 5.2 when evals on Opus at higher effort still fall short.

Canonical: https://routemux.com/anthropic/claude-fable-5-2
Provider: Anthropic

## Key facts

| Field | Value |
| --- | --- |
| Model ID | anthropic/claude-fable-5-2 |
| Official model ID | claude-fable-5-2 |
| Context window | 1M |
| Max output | 128K |
| Input modalities | text, image |
| Output modalities | text |
| Capabilities | Vision, Reasoning, Tool Calling, Coding, Cache, Paid Cache Write |
| Upstream release | 2026-11-09 |
| Availability | available |

## Pricing

- RouteMux input: $2.8 per 1M tokens
- RouteMux output: $14 per 1M tokens
- Official input: $8 per 1M tokens
- Official output: $40 per 1M tokens
- Cache read: $0.07 per 1M tokens
- Cache write: $3.5 per 1M tokens
- Saving vs official: 65%

Billing is prepaid wallet, charged per successful request only.

## API access

Base URL: https://api.routemux.com

Supported protocols:

- anthropic_messages

## What is Claude Fable 5.2

Claude Fable 5.2 was released by Anthropic on November 9, 2026, succeeding Claude Fable 5.1 alongside Claude Mythos 5.2, which offers the same specifications and pricing only to verified organizations (such as the Cyber Verification Program). Anthropic builds Fable 5.2 for demanding reasoning and long-horizon agentic work—agent sessions that run for hours, multistep deep research, and analysis carried through to a finished document, spreadsheet, or deck. Its comparative latency improves from Slower on Fable 5.1 to Moderate, and retirement is not sooner than November 9, 2027. For most workloads, Anthropic recommends starting on Claude Opus 5.5 and moving to Fable 5.2 when evals on Opus at higher effort still fall short.

Fable 5.2 keeps a 1M-token context window (default and maximum), a 128K-token output limit, text and image input, a June 2026 knowledge cutoff, and the Opus 4.7 tokenizer. Adaptive thinking is always on and cannot be disabled, controlled through five effort levels: low, medium, high (default), xhigh, and max. Anthropic lowered official base input and output prices by 20% compared with Fable 5.1 ($8 / $40 vs. $10 / $50 per 1M tokens), while cache reads remain at 2.5% (0.025x) of the base input rate.

Four API changes from Fable 5.1 can break existing integrations: the preserved-thinking prefix check now applies to every account (editing system, tools, or earlier messages before a thinking block returns a 400 unless prefix_mismatch_behavior is set to "drop_block"); only Fable 5.2 and Mythos 5.2 can read Fable 5.2 thinking blocks; the legacy computer_20251124 tool returns a 400 outside Amazon Bedrock (use computer_toolset_20260801 instead); and with the advisor tool, a Fable 5.2 executor requires a Fable 5.2 or Mythos 5.2 advisor. Forced tool use (tool_choice "any" or "tool") and non-default sampling parameters (temperature, top_p, top_k) continue to return a 400 error.

Through RouteMux, Fable 5.2 costs $2.8 per 1M input tokens against Anthropic's $8, $14 per 1M output tokens against $40, and $0.07 per 1M cache-read tokens against $0.2.

## Key capabilities

- **20% lower official price, cache reads at 0.025x** — Anthropic cut base input and output pricing by 20% compared with Fable 5.1 while keeping cache hits and refreshes at 0.025 times (2.5%) the base input rate. Long-horizon agents that repeatedly read large codebases or documents benefit from both reductions.
- **Always-on adaptive thinking with five effort levels** — Adaptive thinking is always on; budget_tokens and thinking disabled both return a 400. Choose from low, medium, high (default), xhigh, and max—start at high and step up to xhigh for your hardest long-running coding and agentic tasks.
- **Universal prefix check and append-only history tools** — Editing system, tools, or messages before an earlier thinking block returns a 400 on every account. Use compaction on demand (compact-2026-09-04 beta) and mid-conversation tool changes (inline-tools-2026-09-15 beta) to keep conversations append-only.
- **Moderate comparative latency** — Anthropic's models overview rates Fable 5.2's comparative latency as Moderate—up from Slower on Fable 5.1 and matching Opus 5.5—while retaining the full 1M-token context window and 128K output ceiling.

## Other variants in this family

- Claude Fable 5 (`anthropic/claude-fable-5`)
- Claude Fable 5.1 (`anthropic/claude-fable-5-1`)

## Sources

- [Anthropic — Claude Fable 5.2 Overview](https://platform.claude.com/docs/en/models/fable-5-2/overview)
- [Anthropic — What's new in Claude Fable 5.2](https://platform.claude.com/docs/en/models/fable-5-2/whats-new-fable-5-2)
- [Anthropic — Introducing Claude Fable 5.2 and Claude Mythos 5.2](https://www.anthropic.com/claude-fable-and-mythos-5-2)
- [Anthropic — Pricing (cache multiplier footnote)](https://platform.claude.com/docs/en/about-claude/pricing)

_Content reviewed by the RouteMux team on 2026-11-11._