Claude Fable 5.2
Language Available Full comparison ↗ anthropic/claude-fable-5-2 · by Anthropic
· decoder-only-transformer
Pricing — 1 offering(s)
About this offering: Fast mode: up to 2.5x faster output at 2x the standard rate ($16 / $80 per MTok input / output), on the Claude API, Claude Models on Google Cloud and Microsoft Foundry (Anthropic hosted); not available with the Batch API.
Input tokens
- $8.00 / 1M tokens (input)
Current
2026-11-09 → present
20% below Fable 5.1's $10. The models-overview table and the 2026-11-09 release notes show the same figure.
Output tokens
- $40.00 / 1M tokens (output)
Current
2026-11-09 → present
20% below Fable 5.1's $50. The models-overview table and the 2026-11-09 release notes show the same figure.
Cache hits and refreshes
- $0.20 / 1M tokens (cached input) Current 2026-11-09 → present
5-minute cache writes
- $10.00 / 1M tokens (cache write) Current 2026-11-09 → present
Batch
Batch input tokens
- $4.00 / 1M tokens (input) Current 2026-11-09 → present
Batch output tokens
- $20.00 / 1M tokens (output) Current 2026-11-09 → present
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how Claude Fable 5.2 fits into a cost-aware routing setup
See how →About Claude Fable 5.2
Claude Fable 5.2 is Anthropic's top generally available model, launched on November 9, 2026 alongside Claude Mythos 5.2, the same underlying model offered only to organizations in Anthropic's verification programs and not tracked here. It succeeds Claude Fable 5.1, which Anthropic now lists as a legacy model. It's a decoder-only transformer with adaptive thinking always on, steered by a five-level `effort` parameter that defaults to high on the Claude API.
List pricing dropped 20% from Fable 5.1, to $8 / $40 per million input / output tokens, with cache reads at $0.20 per million. An optional fast mode speeds output by up to 2.5x for $16 / $80. It has a 1M-token context window and 128K maximum output, and is available on the Claude API, Amazon Web Services, Google Cloud and Microsoft Azure.
Capability profile
How Claude Fable 5.2 rates across core capability dimensions, with the task-level evidence behind each rating.
Anthropic positions Fable 5.2 'for demanding reasoning and long-horizon agentic work' (models overview), and its launch post reports it a step above Opus 5.5 in almost all the capabilities evaluated, including Humanity's Last Exam. Those figures are vendor-reported and not recorded as structured scores here; see limitations.
Launch post: on FrontierCode, Fable 5.2 at High effort scores about level with Opus 5.5's best (45.3% vs 45.5%) for roughly half the cost per task, and reaches 49.1% at Xhigh effort (vendor-reported). Independent: Terminal-Bench 4.0 65.15%, tied first (Vals AI, High effort; 58.59% if the tasks a fallback model answered count as failures). Not yet scored on LiveCodeBench or Terminal-Bench 2.1, the boards Modelglass's coding-task routing ranks on.
Positioned for long-horizon agentic work. Computer use runs through the newer computer-use toolset only (the 2025-11-24 tool returns an error outside Amazon Bedrock), and forced tool choice isn't supported.
1M-token context window and 128K max output (models overview table).
Comparative latency is listed as 'Moderate' in Anthropic's models overview, the same as Opus 5.5 (Fable 5.1 was 'Slower'). The launch post says it generates output up to 30% faster than Fable 5.1 (vendor claim). An optional fast mode raises output speed by up to 2.5x at 2x the price.
$8/$40 per MTok, 20% below Fable 5.1 and 2x Opus 5.5's $4/$20. Cache hits at $0.20/MTok (0.025x base input) keep heavily cached agentic loops cheap. Launch post: at Medium effort it completes challenging tasks with fewer tokens than Fable 5.1 (vendor claim).
Ratings are estimated — limited independent data is available for this model.
Coding benchmarks
View the full coding benchmark leaderboard →| Benchmark | Score | Harness | Source |
|---|---|---|---|
| terminal-bench-4-0 | 65.2% 2026-11 | mini-swe-agent | 📊 leaderboard ↗ |
Operator guidance
Anthropic's top generally available model. Its models overview recommends starting with Claude Opus 5.5 for most workloads and moving to Fable 5.2 only when evals at higher Opus 5.5 effort still fall short, so route here for the hardest reasoning and long-horizon agentic work, not by default. It costs 2x Opus 5.5 per token, and cheap cache reads narrow the gap on heavily cached agentic loops. Replaces Claude Fable 5.1, now legacy, at a lower price; check Anthropic's "What's new in Claude Fable 5.2" breaking changes before switching. Don't route requests that need forced tool choice or thinking disabled.
Use cases
- Demanding reasoning and long-horizon agentic work (Anthropic's own stated positioning)
- Workloads where evals at higher Opus 5.5 effort still fall short (Anthropic's models overview guidance)
- Existing Fable 5.1 workloads, at 20% lower per-token cost, once code is checked for Anthropic's listed breaking changes (edited earlier turns, the old computer-use tool, advisor pairing)
Limitations
- Requires 30-day data retention, like Fable 5.1; not available under zero data retention unless Anthropic expressly authorizes it.