Anthropic and Google Cut Prices on Their Best AI Models
Anthropic cut Claude Fable 5.2 API prices and sped replies; Mythos is restricted, and Google opened Gemini 4 to developers.
Key Takeaways
- Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 with lower API pricing: $8 per million input tokens and $40 per million output tokens.
- Claude Fable 5.2 replies are claimed to be up to 30% faster than Claude 5.1, while an optional fast mode doubles prices but increases speed by up to 2.5x.
- The two models share the same underlying capability but differ in access and safety: Claude Fable 5.2 is public with stronger safeguards, while Claude Mythos 5.2 is restricted to vetted organizations.
- Anthropic reports a self-reported BenchCAD 2.1 score of 83.5 for Claude Fable 5.2, up from 44.4 for Claude Fable 5.1, though this does not guarantee broader performance gains.
- Lower token prices may not always mean lower total costs, so developers should test real workflows; the article also highlights Google’s Gemini 4 Carbon opening to programmers.
For developers who build on Anthropic’s Claude assistant, the week of November 7–13, 2026 brought a practical update. The Real Python AI News weekly edition said big AI companies spent the week making their best models cheaper and faster to use, and it introduced two cases where AI output could be checked line by line. The main change covered here is Claude Fable 5.2, a new version of Anthropic’s most capable AI model, now paired with lower API prices, faster replies, and a higher self-reported score on BenchCAD 2.1.
What Anthropic Changed
Anthropic says Claude Fable 5.2 and Claude Mythos 5.2 are new versions of its most capable AI models, and the two share one underlying model. The difference is access and safety behavior. Claude Fable 5.2 has strong safeguards around cybersecurity and biology, so anyone can use it. Claude Mythos 5.2 has fewer safeguards and is open only to vetted organizations.
Developers connect their applications to Claude through an API, which lets one software program talk to another. They pay by the token, a small piece of text such as a short word. Under the new pricing, sending one million tokens costs $8, and one million tokens of replies cost $40.
Anthropic says replies come up to 30 percent faster than with version 5.1. For teams that need quicker output, an optional fast mode doubles the listed prices to $16 and $80. Anthropic says that mode writes up to 2.5 times faster.
What the Benchmark Number Shows
On BenchCAD 2.1, a hardware engineering test, Anthropic reports a score of 83.5 for Claude Fable 5.2 at its highest effort setting. That is up from 44.4 for Claude Fable 5.1. These are Anthropic’s own measurements, not an outside audit.
The number points to a large self-reported jump on this particular test. It does not by itself prove that every coding, research, or automation job will become cheaper or more reliable. The practical result depends on how much the model must think, write, or revise for a given task.
Why Cheaper Tokens Can Still Mean Costly Jobs
A lower per-token price is easy to understand, but the total bill depends on how many tokens a finished job uses. Philipp Acsany of Real Python framed the new pricing as something that needs testing in real workflows:
“A cheaper token doesn’t always mean a cheaper job. A model that thinks longer or writes more uses more tokens, so a finished task can still cost more. Anthropic says Fable 5.2 needs fewer tokens per task than 5.1 at its default setting. Businesses find out the real cost by trying it on their own work.” — Philipp Acsany, Real Python
That gives developers a clear action: compare Claude Fable 5.2 with Claude Fable 5.1 on representative prompts, measure tokens used per completed task, and decide whether the optional fast mode is worth double the price. If a task finishes with fewer tokens, the lower rates can reduce cost. If the model writes longer answers, the savings may disappear.
Access Limits and Model Choice
The two new models sit on the same underlying capability, but they are not interchangeable for every user. Claude Fable 5.2 is the publicly available option, with stronger safety constraints in sensitive areas. Claude Mythos 5.2 is restricted to vetted organizations and has fewer safeguards.
For most developers, the public choice is Claude Fable 5.2. Its pricing, speed claim, and BenchCAD 2.1 score are the details to test. For organizations that can access Claude Mythos 5.2, the key question is whether the reduced safeguards and restricted access fit their use case. The source material does not give Mythos 5.2 pricing, speed, or benchmark details beyond the shared-model statement.
Other Coverage in the Edition
The Real Python AI News weekly edition also linked to a separate article, “Google’s Most Capable AI Model, Gemini 4 Carbon, Opens to Programmers,” at https://realpython.com/ai-news-weekly-2026-11-13/.
Meta description: Anthropic's Claude Fable 5.2 and Mythos 5.2 get lower API prices, faster replies, and new BenchCAD scores, explained for developers in plain terms.
Tags: Claude Fable 5.2, Anthropic API, AI pricing, BenchCAD 2.1, Real Python AI News
Featured image: abstract, friendly illustration style, no real people, no logos, no text.

Comments (0)