# Anthropic ships Claude Fable 5.2 at a 20 percent lower price and keeps Mythos 5.2 behind vetted access

Fable 5.2 lists at $8 per million input tokens and $40 per million output tokens, down from $10 and $50, and Anthropic's own tables show Mythos 5.2 at 89.6 percent on a CVE-rediscovery test, against 78.3 percent for Mythos 5.1.

- Published: 2026-11-10T06:14:13.273Z
- Canonical: https://polylog.news/ai/2026-11-10/anthropic-ships-claude-fable-5-2-at-a-20-percent-lower-price
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Anthropic News](https://www.anthropic.com/claude-fable-and-mythos-5-2), [Polylog editors](https://polylog.news), [eesel AI](https://www.eesel.ai/blog/claude-mythos-5-2), [DataCamp](https://www.datacamp.com/blog/claude-fable-5-2-and-mythos-5-2), [LLM Stats](https://llm-stats.com/blog/research/claude-fable-5-2-launch), [Anthropic Research: Formalizing Fermat's Last Theorem](https://www.anthropic.com/research/formalizing-fermats-last-theorem)

Anthropic [released Claude Fable 5.2 and Claude Mythos 5.2](https://www.anthropic.com/claude-fable-and-mythos-5-2) on November 9. It calls them its most capable models for coding, finance and the sciences. The split follows the pattern Anthropic set with the version 5 generation. Fable is the general release with safeguards. Mythos is a restricted-access version for organizations that pass Anthropic's cyber and life-sciences verification programs, [according to prior coverage of the Mythos line](https://en.wikipedia.org/wiki/Claude_Mythos).

The most immediate commercial change is price. Several trackers report Fable 5.2 at [$8 per million input tokens and $40 per million output tokens](https://llm-stats.com/blog/research/claude-fable-5-2-launch), down from $10 and $50 for Fable 5.1. They also report a fast mode at [$16 and $80](https://www.datacamp.com/blog/claude-fable-5-2-and-mythos-5-2), a 1M-token context window and 128K maximum output.

On capability, every number available so far comes from Anthropic. Mythos 5.2 scores [89.6 percent on CVE Rediscovery, against 78.3 percent for Mythos 5.1, and 76.3 percent on CyScenarioBench, against 61.7 percent](https://www.eesel.ai/blog/claude-mythos-5-2). On a 24-hour de novo protein-binder design task it scores 85.1 percent, against 79.1 percent. Fable 5.2 reportedly reaches [49.1 percent on FrontierCode 2.0 at its highest effort setting, 69.9 percent on FrontierSWE v2 and 74.2 percent on Humanity's Last Exam with tools](https://www.datacamp.com/blog/claude-fable-5-2-and-mythos-5-2). Outlets disagree on the Terminal-Bench-Science 0.1 figure, and cached copies of Anthropic's table leave several rows blank. The comparison set is Anthropic's own earlier models (Mythos 5.1 and Opus 5.5). There is no head-to-head against OpenAI's or Google's current models on the same harness.

The most striking claim is mathematical. The Russian-language channel AI ML Big Data [summarized Anthropic's post](https://t.me/ai_machinelearning_big_data/11324) as saying that 1,000 agents formalized the proof of Fermat's Last Theorem in 30 hours. Anthropic's [September formalization](https://www.anthropic.com/research/formalizing-fermats-last-theorem) took dozens of agents 11 days, produced about 13 million lines of Lean (a computer proof-checking language) and was checked against Mathlib, Lean's library of formalized mathematics, including its statement of the theorem. The 30-hour result has no published proof artifact yet. It measures wall-clock throughput from parallel agents on a target already solved once. It does not show new mathematical ability.

The release comes days before Anthropic's expected Nasdaq debut. A visible capability improvement and a price cut in the same week strengthen the company's pitch to investors. That timing does not make the numbers inaccurate, but it is a reason to wait for independent verification.

## What this means

The mechanism is price, not just capability. A 20 percent cut on the general-release model lowers the cost of long-horizon coding agents, where output tokens dominate the bill. Pressure falls on OpenAI's and Google's flagship pricing and on resellers whose margin depends on Anthropic list rates. The gated Mythos tier keeps the most commercially sensitive cyber capability inside a vetted customer base. That base becomes a distribution channel governments can influence, because verification decides which firms and countries get access. Markets already expected a model refresh before the initial public offering. The new information is that Anthropic chose to cut price rather than hold margin. That suggests it sees share, not unit economics, as the metric investors will reward. The precedent is OpenAI's repeated price cuts on GPT-4-class models in 2023 and 2024, each of which rivals matched within weeks. The likelier outcome is a matching price move from at least one rival frontier vendor within weeks. That depends on independent evaluators confirming that Fable 5.2's coding gains hold outside Anthropic's harness. If third-party runs on Artificial Analysis or SWE-bench-style leaderboards show the gains shrinking, the cut will look defensive and rivals will hold their prices.

## What to watch

- Whether Anthropic publishes the Lean artifacts for the 30-hour Fermat run. A checkable proof would turn a throughput claim into a reproducible result.
- Independent leaderboard results for Fable 5.2 against OpenAI's and Google's current flagships. Anthropic compared it only with its own models.
- Price changes from OpenAI or Google on their top coding models in the next few weeks, which would show whether this cut starts a new round of repricing.