Anthropic said July 24 that biology-related requests blocked on Claude Fable 5 will now route to Claude Opus 5 rather than Opus 4.8. The company tied the change to Opus 5's safeguards, which it described as similar to Opus 4.8's, and called the new model its most capable generally available model for scientific research. Fable 5 is the model the U.S. government temporarily placed under export controls in June after Amazon researchers reported they could bypass its safeguards, Fortune reported.

Anthropic pulled Fable 5 from the market on June 12 and released it again on June 30 after strengthening its security. Mythos 5, which Anthropic says remains the stronger model for long-running autonomous biological work, has gone only to a select group of organizations and government agencies, Gizmodo reported.

What Changed

AI-generated summary, reviewed by an editor. More on our AI guidelines.

What the behavioral audit measured

Anthropic's launch post called Opus 5 its "most aligned model to date," reporting closer adherence to Claude's Constitution than Opus 4.8, Sonnet 5 or Fable 5, along with the lowest rate of deceptive behavior among the models it compared. The company put the model's score for overall misaligned behavior at 2.3, the lowest of its recent models, and presented that result as a chart. Anthropic described the audit as part of its own pre-deployment testing. The same post called Opus 5 the least susceptible to being tricked into misuse and the safest at avoiding reckless actions that could have hard-to-reverse side effects.

Biology routing shifts to Opus 5

A biology request blocked on Fable 5 now goes to Opus 5, according to the launch post. Requests flagged on Opus 5 inside Claude.ai, Claude Code and Claude Cowork continue to fall back to Opus 4.8 by default, and API customers can switch on automatic fallbacks that route a flagged Opus 5 or Fable 5 request to the best available model.

An Anthropic spokesperson told VentureBeat that the fallback to Opus 4.8 rests on capability, since that model has "lower capability levels making the risk of harmful use lower as well." Users see a message in the chat when a fallback occurs, the spokesperson added.

Anthropic wrote that Opus 5 "does not advance the frontier in risky, dual-use capabilities" and remains behind Mythos 5 in both biology research and offensive cybersecurity, citing evaluations it says it ran with private-sector and government partners. The company has also said it deliberately kept cyber training out of the model. Its launch post identifies long-running autonomous research as the area where AI models could pose the most substantial biology-related risks, and the spokesperson named autonomous drug-design campaigns as work where Mythos 5 stays stronger.

Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.

The internal life-sciences results

Opus 5 beat Opus 4.8 on every one of Anthropic's internal life-sciences evaluations, the company reported. Its largest chemistry gain came from inferring molecular structures from spectroscopy data, 10.2 percentage points above Opus 4.8 on an internal benchmark. Internal protein tasks, which test how changes in a sequence affect function, improved by 7.7 points. SiliconANGLE noted that automated protein research was a focus of the 2024 Nobel Prize in chemistry.

Anthropic has separately said Opus 5 "overperforms on biology tasks compared to Opus 4.8." ZDNET wrote that the company had not defined what "overperformed" means in that context. Neither ZDNET nor SiliconANGLE tested the results independently.

Opus 5 prices at half Fable 5's rates

Opus 5 reached all Anthropic platforms on July 24 at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8. The Register lists Fable 5 at $10 per million input tokens and $50 per million output tokens, so a biology request rerouted off Fable 5 now lands on a model priced at half those rates. General access to Opus 5 carries no data-retention requirement, according to Anthropic, and the release is the company's fourth in less than two months, following Mythos 5, Fable 5 and Sonnet 5.

Frequently Asked Questions

What exactly did Anthropic change about biology requests?

Biology-related requests that are blocked on Claude Fable 5 now route to Claude Opus 5 instead of Opus 4.8, Anthropic said in its July 24 launch post. The company tied the change to Opus 5's safeguards, which it described as similar to Opus 4.8's.

What is the 2.3 score Anthropic cites?

It is Opus 5's score for overall misaligned behavior on Anthropic's automated behavioral audit, which the company says is the lowest among its recent models. Anthropic presented the result as a chart and described the audit as part of its own pre-deployment testing.

Is Opus 5 now Anthropic's strongest model for biology work?

Anthropic calls it the most capable generally available model for scientific research, but says Mythos 5 remains stronger for long-running autonomous biological work. A spokesperson named autonomous drug-design campaigns as work where Mythos 5 stays ahead.

Why was Fable 5 under export controls?

The U.S. government temporarily placed Fable 5 under export controls after Amazon researchers reported they could bypass its safeguards, Fortune reported. Anthropic pulled the model on June 12 and released it again on June 30 after strengthening its security.

What does Opus 5 cost?

Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8. The Register lists Fable 5 at $10 and $50 per million tokens, so a rerouted biology request lands on a model priced at half those rates.

AI-generated summary, reviewed by an editor. More on our AI guidelines.

Trump Officials Revive Push to Bar Chinese AI Models After Kimi K3
Four separate attempts to restrict Chinese AI models reached internal consideration inside the Trump administration last year and were killed before any took effect, Axios reported Monday, and parts o
The Implicator
Anthropic Ships Fable 5 and Locks the Unrestricted Version Away
San Francisco | Wednesday, June 10, 2026 Anthropic put a Mythos-class model on the open market Tuesday. Claude Fable 5 runs $10 per million input tokens and $50 out, and its classifiers hand any cybe
The Implicator
The White House Is Asking Anthropic for the Impossible
A government can stop a risky AI model from reaching foreign users. But perfect jailbreak resistance is not a compliance standard any lab can reliably meet, and that is the direction officials now app
The Implicator
AI News

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai