- Context
- 1Mtokens
- Listed
- 5d ago
Cost/1M tokens
Anthropic's Claude Fable 5.2 is an upgrade of Claude Fable 5.1 and shares an underlying model with Claude Mythos 5.2, with cybersecurity and biology safeguards that let Anthropic offer it to all users. Anthropic says it generates output up to 30% faster than Fable 5.1 and uses fewer tokens on challenging tasks. It is built for demanding, open-ended technical work such as coding, agentic tasks, and research.
- Multimodal model
- Accepts images, files
- 1M context
- Reasoning
- Tools
- Structured output
How it compares
Score vs. costArena WebDev
Each model at its best score · Dashed: the best score at each cost
- Anthropic
- Google DeepMind
- OpenAI
- Z.ai
- Meta
- xAI
- Moonshot AI
- Xiaomi
PremiumA premium line the lab sells above its own flagship, for the hardest work at the highest price.
Premium models from other labs
Add the models you use to see how this one compares.
Sign inAdditional source evidence
- Source model
- anthropic/claude-fable-5-2
- Source family
- claude-fable
- Last updated
- 2026-11-09
- Knowledge cutoff
- 2026-06
- Model context
- 1M
- Model output limit
- 128K
- Open weights
- No
- File attachments
- Yes
- Reasoning
- Yes
- Tool calling
- Yes
- Temperature control
- No
Published benchmarks
| Benchmark | Score | Evaluation details | Evidence |
|---|---|---|---|
| FrontierCode | 49.1 · score | xhigh effort; safeguards on · 2.0 · 2026-11-09 | Source |
| FrontierSWE | 69.9 · score | highest available effort; safeguards on · 2 · 2026-11-09 | Source |
| AutomationBench | 51.4 · score | highest available effort; safeguards on · 2026-11-09 | Source |
| Terminal-Bench-Science | 74.6 · score | highest available effort; safeguards on · 0.1 · 2026-11-09 | Source |
| WANDR | 82.1 · relaxed score | highest available effort; safeguards on · 2026-11-09 | Source |
| OSWorld | 89.3 · partial score | offline set · highest available effort; safeguards on · 2.1 · 2026-11-09 | Source |
| RoboDojo Sim | 64.9 · average success rate | max effort; internal run · 2026-11-09 | Source |
| AA-Briefcase | 66.1 · rubric score | highest available effort; safeguards on · 1.1 · 2026-11-09 | Source |
| AA-Briefcase | 1785 · Elo | highest available effort; safeguards on · 1.1 · 2026-11-09 | Source |
| Remote Labor Index | 41.3 · score | high effort; safeguards on · 2026-11-09 | Source |
| Humanity's Last Exam | 74.2 · score | with tools; highest available effort; safeguards on · 2026-11-09 | Source |
| Chartography | 73.4 · score | no tools; highest available effort; safeguards on · 2026-11-09 | Source |
Complete source record (JSON)
{
"id": "anthropic/claude-fable-5-2",
"name": "Claude Fable 5.2",
"description": "Claude model for demanding reasoning and long-horizon agentic work",
"family": "claude-fable",
"last_updated": "2026-11-09",
"knowledge": "2026-06",
"attachment": true,
"reasoning": true,
"tool_call": true,
"temperature": false,
"modalities": {
"input": [
"text",
"image",
"pdf"
],
"output": [
"text"
]
},
"limit": {
"context": 1000000,
"output": 128000
},
"benchmarks": [
{
"name": "FrontierCode",
"score": 49.1,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "score",
"variant": "xhigh effort; safeguards on",
"version": "2.0",
"date": "2026-11-09"
},
{
"name": "FrontierSWE",
"score": 69.9,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "score",
"variant": "highest available effort; safeguards on",
"version": "2",
"date": "2026-11-09"
},
{
"name": "AutomationBench",
"score": 51.4,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "score",
"variant": "highest available effort; safeguards on",
"date": "2026-11-09"
},
{
"name": "Terminal-Bench-Science",
"score": 74.6,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "score",
"variant": "highest available effort; safeguards on",
"version": "0.1",
"date": "2026-11-09"
},
{
"name": "WANDR",
"score": 82.1,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "relaxed score",
"variant": "highest available effort; safeguards on",
"date": "2026-11-09"
},
{
"name": "OSWorld",
"score": 89.3,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "partial score",
"variant": "highest available effort; safeguards on",
"dataset": "offline set",
"version": "2.1",
"date": "2026-11-09"
},
{
"name": "RoboDojo Sim",
"score": 64.9,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "average success rate",
"variant": "max effort; internal run",
"date": "2026-11-09"
},
{
"name": "AA-Briefcase",
"score": 66.1,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "rubric score",
"variant": "highest available effort; safeguards on",
"version": "1.1",
"date": "2026-11-09"
},
{
"name": "AA-Briefcase",
"score": 1785,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "Elo",
"variant": "highest available effort; safeguards on",
"version": "1.1",
"date": "2026-11-09"
},
{
"name": "Remote Labor Index",
"score": 41.3,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "score",
"variant": "high effort; safeguards on",
"date": "2026-11-09"
},
{
"name": "Humanity's Last Exam",
"score": 74.2,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "score",
"variant": "with tools; highest available effort; safeguards on",
"date": "2026-11-09"
},
{
"name": "Chartography",
"score": 73.4,
"source": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
"metric": "score",
"variant": "no tools; highest available effort; safeguards on",
"date": "2026-11-09"
}
],
"release_date": "2026-11-09",
"open_weights": false
}Text provenance
Watermarked · GlobalClaude text watermarking
Anthropic says newer Claude models include an invisible statistical watermark globally. It changes token selection rather than adding hidden characters.
Official explanationFrequently asked questions
API and pricing
Claude Fable 5.2 costs $8.00 per 1M input tokens and $40.00 per 1M output tokens. Across 5 paid providers, input prices range from $8.00 to $16.00 per 1M tokens.
Amazon Bedrock is the cheapest provider at the lowest blended price (3:1 input to output). It charges $8.00 per 1M input tokens and $40.00 per 1M output tokens.
Yes, prompt caching is supported. Cached input costs $0.20 per 1M tokens, a 98% discount from the uncached $8.00 per 1M tokens, and cache writes cost $10.00 per 1M tokens.
No, the batch input discount versus standard is 0%. Amazon Bedrock offers batch pricing at $8.00 per 1M input tokens and $40.00 per 1M output tokens.
A typical request of 10,000 input tokens and 1,000 output tokens costs $0.12 at the cheapest provider, Amazon Bedrock.
In the last day, 4 endpoints reported uptime data and all 4 were at 99% uptime or better. The best uptime was 100.0%, by Anthropic, Azure and Google.
Capabilities and specs
The context window is 1M tokens (1,000,000). Maximum output is 128K tokens (128,000).
Claude Fable 5.2 supports text, image and file inputs. Audio and video are not listed among its input modalities.
Yes, tool calling (function calling) is supported.
Yes, structured output (JSON schema or JSON mode) is supported.
Yes, reasoning is supported and is always on. Effort levels are max, xhigh, high, medium and low, with high as the default. All 5 tracked providers expose reasoning, and a provider offers the reasoning_effort setting.
The knowledge cutoff of Claude Fable 5.2 is 2026-06.
Quality and benchmarks
Claude Fable 5.2 ranks 1 on Arena WebDev with a score of 1836. This is Arena's own rank.
Yes, Claude Fable 5.2 is ahead of Claude Fable 5.1 on ARC-AGI-2, Arena WebDev, FrontierMath-Tier-4-v2-Private, FrontierMath-Tiers-1-3-v2-Private and SimpleQA Verified, and behind on OTIS Mock AIME 2024-2025.
GPT-6 Astra by OpenAI, Fugu Ultra v2 by Sakana AI and Muse Wonder by Meta, matched on model kind, blended price and release date. Their blended prices are $20.00, $11.25 and $8.00 per 1M tokens respectively.
GPT-6 AstraClaude Fable 5.2 vs GPT-6 AstraFugu Ultra v2Claude Fable 5.2 vs Fugu Ultra v2Muse WonderClaude Fable 5.2 vs Muse Wonder
Kimi K3 is the highest-ranked open-weight model on Arena Text, at rank 19. Claude Fable 5.2 itself is closed.
Licensing and access
No, Claude Fable 5.2 is closed.
Yes, Claude Fable 5.2 is available on Azure, Amazon Bedrock and Google Vertex AI.
Basics
Claude Fable 5.2 was made by Anthropic and was first listed on OpenRouter on November 7, 2026.
No, the listed API price is $8.00 per 1M input tokens and $40.00 per 1M output tokens.
Generated from Vertical Tabs data, updated November 11, 2026.
Discussion
No comments yet
Be the first to share your thoughts.