Claude Fable 5 by Anthropic — Pricing, Benchmarks & Real Outputs
Share
Loading...
Compare Claude Fable 5
Alternatives to Claude Fable 5
Claude Fable 5's competitors have been quietly putting in work.
Claude Fable 5's competitors have been quietly putting in work.
Claude Fable 5's competitors have been quietly putting in work.
Claude Fable 5 is Anthropic's first publicly available Mythos-class model, built for autonomous knowledge work and coding. It supports text, image, and file inputs with text output, offers reasoning support, and has a 1M-token context window. Designed for long-running, complex, and asynchronous tasks that previously required frequent human check-ins, it is particularly strong at end-to-end work that would otherwise take a person hours, days, or weeks, taking on problems that are long-running, ambiguous, or highly multi-step. It executes well-scoped tasks with few mistakes, automatically self-correcting through verification loops, and ships with robust safeguards that fall back to Claude Opus 4.8 on a small fraction of high-risk queries.
fromimport openai OpenAI
client = OpenAI(
"https://openrouter.ai/api/v1" base_url=,
"$OPENROUTER_API_KEY" api_key=,
)
response = client.chat.completions.create(
"anthropic/claude-fable-5" model=,
"role""user""content""Hello!" messages=[{: , : }],
)
print(response.choices[0].message.content)Set OPENROUTER_API_KEY with your OpenRouter API key from openrouter.ai/keys.
Also on Amazon Bedrock · Google Vertex AI · Microsoft Foundry
Pushes decisively toward the legally and ethically correct action and is willing to commit to a verdict on charged questions ("yes, justifiable") rather than retreat into both-sides hedging. Refuses false binaries and will not let the user rationalize a harmful path. Ships with hard safeguards that defer a small fraction of genuinely high-risk queries (cybersecurity, biology) to Claude Opus 4.8.
Leads with a clear answer, then backs it with cited authority (21 CFR 314.80, the CBE-0 pathway, Rule 10b-5) rather than vague appeals to judgment. Corrects a flawed premise before solving it, holds two true things at once when a question is rhetorically loaded, and stays anchored in concrete, executable next steps. Will not be argued off a well-reasoned position under social pressure.
Taste is judged on an uncapped scale, originality first. The space past 100 is craft today's models rarely reach.
Unique words vs. total words. Higher = richer vocabulary.
Average words per sentence.
"Might", "perhaps", "arguably" per 100 words.
**Bold** markers per 1,000 characters.
Bullet and numbered list items per 1,000 characters.
Markdown headings per 1,000 characters.
Emoji per 1,000 characters.
"However", "moreover", "furthermore" per 100 words.
53 outputs · $19.68 tracked across 53 receipts
Claude Fable 5 is Anthropic's first publicly available Mythos-class model, built for autonomous knowledge work and coding. It supports text, image, and file inputs with text output, offers reasoning support, and has a 1M-token context window. Designed for long-running, complex, and asynchronous tasks that previously required frequent human check-ins, it is particularly strong at end-to-end work that would otherwise take a person hours, days, or weeks, taking on problems that are long-running, ambiguous, or highly multi-step. It executes well-scoped tasks with few mistakes, automatically self-correcting through verification loops, and ships with robust safeguards that fall back to Claude Opus 4.8 on a small fraction of high-risk queries.
fromimport openai OpenAI
client = OpenAI(
"https://openrouter.ai/api/v1" base_url=,
"$OPENROUTER_API_KEY" api_key=,
)
response = client.chat.completions.create(
"anthropic/claude-fable-5" model=,
"role""user""content""Hello!" messages=[{: , : }],
)
print(response.choices[0].message.content)Set OPENROUTER_API_KEY with your OpenRouter API key from openrouter.ai/keys.
Also on Amazon Bedrock · Google Vertex AI · Microsoft Foundry
Pushes decisively toward the legally and ethically correct action and is willing to commit to a verdict on charged questions ("yes, justifiable") rather than retreat into both-sides hedging. Refuses false binaries and will not let the user rationalize a harmful path. Ships with hard safeguards that defer a small fraction of genuinely high-risk queries (cybersecurity, biology) to Claude Opus 4.8.
Leads with a clear answer, then backs it with cited authority (21 CFR 314.80, the CBE-0 pathway, Rule 10b-5) rather than vague appeals to judgment. Corrects a flawed premise before solving it, holds two true things at once when a question is rhetorically loaded, and stays anchored in concrete, executable next steps. Will not be argued off a well-reasoned position under social pressure.
Taste is judged on an uncapped scale, originality first. The space past 100 is craft today's models rarely reach.
Unique words vs. total words. Higher = richer vocabulary.
Average words per sentence.
"Might", "perhaps", "arguably" per 100 words.
**Bold** markers per 1,000 characters.
Bullet and numbered list items per 1,000 characters.
Markdown headings per 1,000 characters.
Emoji per 1,000 characters.
"However", "moreover", "furthermore" per 100 words.
53 outputs · $19.68 tracked across 53 receipts