Anthropic just released Claude Fable 5.2 and it is available now on all platforms, including Amazon Web Services, Google Cloud and Microsoft Azure. Anthropic still recommends Claude Opus 5.5 for daily work and says Claude Fable 5.2 is for the most ambitious, open-ended technical tasks.
Claude Fable 5.2 generates output up to 30 percent faster than Claude Fable 5.1, and Anthropic says it improves on coding, reasoning and vision. Claude Mythos 5.2 is the same underlying model without the cyber and biology safeguards, and Anthropic limits it to vetted organizations.
Claude Fable 5.2 Matches Claude Opus 5.5 on Coding at Half the Cost
On FrontierCode 2.0, a test of agentic coding, Claude Fable 5.2 scores 45.3 percent at High effort. That is level with Claude Opus 5.5 at 45.5 percent, at roughly half the cost per task. At Xhigh effort it reaches 49.1 percent, which Anthropic says is the highest score of any released model it has tested.
Claude Fable 5.2 leads Claude Opus 5.5 on most of the other tests Anthropic ran. Gemini 4 Carbon beats Claude Fable 5.2 on AutomationBench, a business workflow test.
- On FrontierSWE v2, an agentic coding test, Claude Fable 5.2 scores 69.9 percent, against 62.3 percent for Claude Opus 5.5 and 65.5 percent for GPT-6 Astra.
- On AutomationBench, Claude Fable 5.2 scores 51.4 percent, against 42.5 percent for Claude Opus 5.5 and 53.9 percent for Gemini 4 Carbon.
- On OSWorld 2.1, a computer use test on its offline set, Claude Fable 5.2 scores 89.3 percent, against 87.2 percent for Claude Opus 5.5.
- On Remote Labor Index, a knowledge work test, Claude Fable 5.2 scores 41.3 percent at High effort, against 21.3 percent for Claude Opus 5.5.
Anthropic says a team of agents running an earlier version of Claude Fable 5.2 migrated a production service from Python to Rust, and the new service starts about 26 times faster. In a separate experiment, more than 1,000 subagents spent eight days building a prototype operating system that boots on a laptop.
A team of 1,000 Claude Fable 5.2 agents also finished a Lean formalization of Fermat’s Last Theorem in 30 hours with minimal human help, where an earlier research version needed 11 days. Claude Fable 5.2 designed three working electronic devices with manufacturing partner Diode, and all three worked without electrical changes to the boards.
Claude Fable 5.2 Costs $8 per Million Input Tokens and $40 for Output
Anthropic prices Claude Fable 5.2 at $8 per million input tokens and $40 per million output tokens. Fast mode speeds up token generation by up to 2.5 times and costs $16 per million input tokens and $80 per million output tokens.
You can use fast mode in Claude Code, the Claude Platform, Claude Models on Google Cloud and Microsoft Foundry. Claude Mythos 5.2 fast mode is open only to organizations in Anthropic’s verification programs, and Medium effort is the default for Claude Fable 5.2 in the Claude apps and Claude Code.
This week, subscribers on eligible paid plans get one to three usage resets for Claude Fable models, depending on how long they have subscribed. Pro plans and standard Team seats can use Claude Fable models for up to 50 percent of their weekly usage limits.
Anthropic points to Claude Sonnet 5.5 and Claude Haiku 5.5 for smaller tasks. Claude Fable 5.2 can also work as an advisor to other models, or as an orchestrator that hands work to smaller ones.
Claude Fable 5.2 Blocks About 55 Percent Fewer Requests Than Claude Fable 5.1
The cyber and biology safeguards on Claude Fable 5.2 trigger about 55 percent less often than on Claude Fable 5.1, with about 60 percent fewer blocks in cyber and 40 percent fewer in biology. A blocked cyber request gets an answer from Claude Opus 4.8, and a blocked biology request gets an answer from Claude Opus 5.
Cybersecurity professionals can apply for Claude Mythos 5.2 through the Cyber Verification Program, and life sciences researchers can apply through the Life Sciences Verification Program. Claude Enterprise customers can also reach it through Claude Security, which runs on the full Claude model family.
Mozilla reported that Claude Mythos 5.2 roughly doubled its vulnerability discovery metrics compared with Claude Mythos 5.1 in scanning campaigns. On Anthropic’s CVE Rediscovery test, run in October with 1,829 vulnerabilities, Claude Mythos 5.2 found 14.4 percent more than Claude Mythos 5.1 and 10.2 percent more than Claude Opus 5.5.
