For Developers/Models/Claude Fable 5
Closed SourceAnthropicReleased 2026-06

Claude Fable 5

Anthropic's June 2026 Fable model, now legacy; Anthropic recommends migrating to Fable 5.2 ($8/$40)

Anthropic has replaced this model with Claude Fable 5.2.

Context window

1M

Input / 1M tokens

$10.00

Output / 1M tokens

$50.00

Provider

Anthropic

Priced at $10/$50 per 1M input/output tokens, less than half the price of the earlier Claude Mythos Preview. Cache reads cost $1 per 1M, versus $0.25 on Claude Fable 5.1 (same base price) and $0.20 on Claude Fable 5.2 ($8/$40). Uses the Opus 4.7-era tokenizer (~30% more tokens than pre-4.7 models for the same text). 1M context; 128k max output. · Data verified 2026-11-11

Claude Fable 5 (June 2026) is a Mythos-class model made safe for general use and was Anthropic's most capable generally available model at launch, showing state-of-the-art results on nearly all tested benchmarks at the time - reportedly ~95% on SWE-bench Verified. Its lead over other Claude models grows the longer and more complex the task. At launch it shipped with real-time classifiers that redirected a small share of sensitive queries (cyber, bio/chem, distillation) to Opus 4.8; Anthropic stated more than 95% of sessions involved no fallback. Adaptive thinking is always on. Claude Fable 5.1 replaced it in September 2026, and Claude Fable 5.2 (November 9, 2026, $8/$40) is now the current Fable model; Fable 5 stays in the API as a legacy model (retirement not before June 9, 2027).

Capability index

Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.

Coding
99
Reasoning
98
Math
96
Multimodal
92
Long context
95
Speed
55
Cost efficiency
55

How to access it

Still served on the Claude API (claude-fable-5), Claude Platform on AWS, Amazon Bedrock (anthropic.claude-fable-5), Google Cloud and Microsoft Foundry. Anthropic lists it as a legacy model (retirement not before June 9, 2027) and recommends migrating to Claude Fable 5.2. Claude Mythos 5, which shares its specifications and pricing, is limited to organizations verified through Anthropic's access programs.

Strengths

  • ✓State-of-the-art on nearly all tested benchmarks at its June 2026 launch (~95% SWE-bench Verified)
  • ✓Exceptional long-horizon agentic performance - lead widens on longer, harder tasks
  • ✓Frontier software engineering, scientific research, knowledge work, and vision
  • ✓1M context with 128k output and always-on adaptive thinking
  • ✓Top scores on external evals (Hebbia Finance, Cognition FrontierCode, ~92.6% GPQA Diamond)

Best for developers who...

Long-running autonomous agents on complex, high-value tasksFrontier software engineering and hard multi-service implementationScientific research and advanced analyticsSenior-level financial and legal reasoningAny task where maximum accuracy outweighs cost

When to choose it (and when not to)

Reach for Claude Fable 5 when...

  • →You have long-running agents tuned to Fable 5 and want stable behavior while you test Fable 5.2
  • →Your harness uses forced tool_choice (any or tool), which returns a 400 on Fable 5.1 and Fable 5.2
  • →You need Priority Tier, which Fable 5 supports and Fable 5.1 and 5.2 do not

Look elsewhere if...

  • ✕New projects: Fable 5.2 costs less ($8/$40 vs $10/$50), has cheaper cache reads ($0.20 vs $1 per 1M) and scores higher on Anthropic's benchmarks
  • ✕You are cost-sensitive - at $10/$50 it is in the priciest Claude tier; Opus 5.5 ($4/$20) or Sonnet 5.5 ($2/$10) cost far less
  • ✕Your workloads touch topics that trigger the safety fallback (some sessions get a response from another model)

How to use it

  • ›Reserve it for hard, long-horizon tasks where its lead over Opus 5.5 and Sonnet 5.5 justifies the price
  • ›Adaptive thinking is always on - use the effort parameter to tune depth rather than toggling thinking
  • ›Handle stop_reason:'refusal' and fallback responses gracefully in production
  • ›Recount token budgets under the Opus 4.7-era tokenizer (~30% more tokens)

Quickstart

Python
import anthropic

client = anthropic.Anthropic()

response = client.messages.create(
    model="claude-fable-5",
    max_tokens=16000,
    output_config={"effort": "high"},
    messages=[
        {"role": "user", "content": "Implement this multi-service feature end to end."}
    ],
)
print(response.content[0].text)

Adaptive thinking is always on. A small share of sensitive queries may be answered by Claude Opus 4.8 instead; check stop_details/refusals in production.

API model id: claude-fable-5

Benchmarks

BenchmarkScoreNotes
SWE-bench Verified95.0%Per secondary benchmark aggregators (MorphLLM); Anthropic reports state-of-the-art but publishes charts, not this exact figure
Terminal-Bench 2.188.0%
GPQA Diamond92.6%
SWE-bench Pro80.3%

Source: MorphLLM Claude benchmark table + Anthropic announcement

Compare Claude Fable 5

Compare Claude Fable 5 with any other model

Build a comparison →
All model comparisons →