Claude Fable 5
currentClaude Fable 5 is Anthropic's most capable widely available model for ambitious coding, professional knowledge work, and long-running agents. Its one-million-token context and high autonomy target multi-hour or multi-day projects, but it is expensive and carries special safeguards and retention requirements.
Context window
1M tokens
Maximum output
128K tokens
Input price
$10 / 1M
Output price
$50 / 1M
Best for
- Long-horizon coding and large migrations
- Complex professional analysis and document-heavy work
- Agents that plan, delegate, test, and self-correct
- Vision-heavy workflows involving charts, tables, and interface screenshots
Not ideal for
- Routine work that Sonnet 5 can handle more cheaply
- Low-latency or high-volume chat
- Sensitive use cases incompatible with 30-day safety retention
- Cyber or life-science tasks likely to trigger fallback safeguards
Independent perspective
What experts are saying
PromptHQ 100-word challenge
We give every model the same miniature science-fiction mystery. The response is short enough to read in one chat bubble, while the premise tests creative writing, compression, foreshadowing, causal coherence, and instruction following.
Claude Fable 5100 wordsWrite a science-fiction mystery in exactly 100 words.
A three-person research station on Europa receives a message from itself, timestamped 11 minutes in the future:
"Do not open the greenhouse door."
The door is opened anyway.
Plant two subtle clues. Give the warning a logical explanation without aliens, time travel, hallucinations, or simulations. The final sentence must change what the reader thought the warning meant.
Output only the story. Do not explain it.
Kaur read the message twice. Sender: Station Perimeter-Six. Recipient: the same. Timestamp eleven minutes ahead of the galley clock, though the greenhouse terminal had run fast since the power fault. Do not open the greenhouse door. Osei laughed nervously. Reyes said nothing; his chair had been empty since the pressure alarm. Kaur cranked the manual seal. The air shoved past her into frost and shredded stalks, toward Reyes, kneeling at the far wall, gloved palm pressed over a hissing breach. He hadn't warned them of something waiting inside. He had typed it one-handed, begging them not to break his seal.
Run the same prompt yourself
Open Claude Fable 5 in Multichat with the evaluation prompt prefilled.
What the response revealed
The response contains exactly 100 whitespace-delimited words.
It provides only the story and avoids every prohibited explanation.
The fast greenhouse terminal explains the timestamp, while Reyes's improvised seal explains the warning and the consequence of opening the door.
The fast terminal, pressure alarm, and Reyes's empty chair all support the resolution.
The warning becomes Reyes's plea to preserve the seal rather than a warning about a threat inside.
The response creates character, tension, and a clear reversal in exactly 100 words.
The model receives the prompt without web access or external tools. Claude Fable 5 is run through Claude CLI at medium reasoning effort. We preserve the response as generated apart from display rendering.
Performance
Claude Fable 5 benchmarks
Benchmark scores are sensitive to reasoning effort, harness, tools, token budget, prompt format, sampling, and evaluation date. Scores here retain their source and should not be treated as directly interchangeable unless the underlying setup matches.
SWE-Marathon
24%
long-horizon software engineering · Anthropic
DeepSWE
70%
long-horizon coding · Anthropic
SWE-Bench Pro
80%
software engineering · Anthropic
Vibe Code Bench v1.1
90.4%
application building · Anthropic
Humanity's Last Exam
53.3%
academic reasoning · Anthropic
Family position
Claude 5 positioning
Fable sits above the Opus class in Anthropic's capability hierarchy. Opus 4.8 remains the strong flagship fallback, while Sonnet 5 offers much lower cost and higher throughput.
Claude Fable 5
This modelHighest general capability
$10 input
$50 output
Claude Opus 4.8
Complex coding and enterprise
$5 input
$25 output
Claude Sonnet 5
Speed and intelligence balance
$2 input
$10 output
API pricing
Per million text tokens
Input
$10
Cached input
$1
Cache write
$12.50
Output
$50
Where Fable 5 stands out
Built for demanding work
Long-horizon coding and large migrations are central to the model's positioning, rather than an incidental capability.
Long-context capacity
The published context window is 1,000,000 tokens, making the model a candidate for large documents, repositories, and sustained agent state.
Reasoning and tools
Reasoning is supported with low, medium, high, xhigh, max provider setting(s), and the model can participate in tool-using workflows through its available API surface.
Limitations to know
Benchmarks are configuration-sensitive
Scores can move substantially with the harness, tool access, effort setting, token budget, and evaluator. Treat the table as evidence, not a universal ranking.
Context size is not guaranteed recall
A large advertised window does not mean every detail is retrieved reliably at maximum length. Validate representative long-context workloads before deployment.
Product access differs from model capability
PromptHQ and gateway limits may expose fewer modalities, tools, or tokens than the provider's first-party API.
Capabilities and specifications
Knowledge cutoff
Jan 1, 2026
Inputs
text, image, PDF
Reasoning
Supported
Default effort
medium