Anthropic Ships Claude Fable 5.2 at $8 per Million Input Tokens, Down From $10

By news editor

Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, its top-tier models, at a lower price than their predecessors: $8 per million input tokens and $40 per million output tokens, according to the company’s announcement. InfoWorld reports the previous models were priced at $10 and $50, a cut of 20%.

The two names cover one model. Anthropic says Fable and Mythos share the same underlying model; Fable carries cybersecurity and biology safeguards that let the company sell it to all users, while Mythos, with those safeguards off, is available only to organizations vetted through its Cyber Verification and Life Sciences Verification Programs. An optional fast mode, which Anthropic says speeds token generation by up to 2.5 times, costs $16 per million input tokens and $80 per million output tokens.

For enterprise buyers, the pricing sits inside a tiering the company is explicit about. Anthropic recommends that most coding and knowledge work start with Claude Opus 5.5, which it calls its strongest daily driver, and move up to Fable or Mythos only for the most ambitious, open-ended technical work. The announcement says Fable 5.2 generates output up to 30% faster than Fable 5.1 and can complete challenging tasks with fewer tokens at the default Medium effort setting.

The benchmark claims are Anthropic’s own and have not been independently replicated. On FrontierCode 2.0, which measures whether open-source maintainers would merge a model’s code changes, the company reports 49.1% for Fable 5.2 at its Xhigh effort setting against 32.4% for Fable 5.1 and 45.5% for Opus 5.5. At High effort, Anthropic says Fable 5.2 scores about level with Opus 5.5’s best result for roughly half the cost per task. On its internal CVE Rediscovery evaluation, run in October 2026 against 1,829 published vulnerabilities, the company says Mythos 5.2 found 10.2% more vulnerabilities than Opus 5.5 and 14.4% more than Mythos 5.1.

Analysts interviewed by InfoWorld drew a line between cheaper tokens and cheaper outcomes. “Anthropic has cut the price of a token, not the price of a thought,” said Sanchit Vir Gogia, chief analyst at Greyhound Research, who urged CIOs to budget for effort settings and measure cost against completed tasks. Lian Jye Su, chief analyst at Omdia, told InfoWorld the combination of higher completion rates and lower prices could make long-running agents viable for software migrations, with a caveat: “Reliability remains an issue without human supervision,” Su said.

Anthropic’s own examples of that long-running work come from inside the company. It says a team of agents running an earlier version of Fable 5.2 migrated a production service from Python to Rust, that one engineer had the model coordinate more than 1,000 subagents over eight days to produce a prototype operating system, and that over 90% of changes merged into its core codebase are now written by Claude with human guidance. Named testers quoted in the announcement include Vercel, Shopify, Datadog, Uber, Oracle and Snowflake, but Anthropic gives no customer counts, revenue figures or contract terms for the new models.

The safeguards carry a cost of their own that the company says it has reduced. Fable 5.2 blocks high-risk cyber tasks such as exploit generation and most research biology, handing those queries to older Opus models instead; Anthropic reports about 55% fewer blocks than Fable 5.1 in early testing. On data handling, customers eligible for zero data retention can keep it, while others face 30-day retention for safety monitoring. A new Enterprise Frontier Safeguards option, which runs monitoring on prompts stored in the customer’s own infrastructure, is in private beta with Bridgewater, Salesforce and Stripe, the announcement says. Gogia told InfoWorld that retention requirements could outweigh the price cut for regulated enterprises and advised procurement teams to get zero-retention terms in writing.

Fable 5.2 is available on Amazon Web Services, Google Cloud and Microsoft Azure as well as Anthropic’s own platform. The next thing to watch is whether third-party evaluators publish results on the new models, and whether the per-task cost savings Anthropic claims at High effort show up in customers’ bills once agents run unattended for hours.