header-langage
简体中文
繁體中文
English
Tiếng Việt
한국어
日本語
ภาษาไทย
Türkçe
Scan to Download the APP
Partnered With:

Anthropic Unveils Most Powerful Mythos-Level Model Fable5: Programming Performance Far Exceeds All Prior Models

According to TrendWatch Beating Monitor, Anthropic has launched the all-new flagship large-scale model Claude Fable 5, the first publicly available Mythos-level model. The new model is priced at $10 per million input tokens and $50 per output, a price reduction of over half compared to the test version. Targeted at security agencies, the security-restricted upgrade Claude Mythos 5 of the same architecture has also been released simultaneously.

In FrontierCode's Diamond difficulty evaluation, Claude Fable 5 scored 29.3%, far surpassing GPT-5.5 at 5.7% and Claude Opus 4.8 at 13.4%. In a test by Stripe, the model completed the migration of a 50 million-line Ruby codebase in one day, a task that would have taken a team two months in the past.

In terms of visual capabilities, the new model can directly pass through the game "Pokémon FireRed" with a screen capture and rebuild webpage source code. In the field of biology, the model can predict Dyno Therapeutics' adenovirus capsid assembly through inference alone, outperforming dedicated protein models. The new model has also independently conducted single-cell genomics research, with a designed and trained classification model superior to a similar model published in Science magazine, and with only one hundredth of the parameters.

To prevent the misuse of cutting-edge capabilities, Anthropic has deployed an independent security classifier that will automatically fallback to Claude Opus 4.8 when encountering high-risk queries, affecting less than 5% of sessions on average. To guard against jailbreak attacks, enterprise traffic data will be retained for 30 days, and all manual reviews during this period will be logged. For Glasswing partners, the network security restrictions of Claude Mythos 5 have been lifted, and the biomedicine restrictions will gradually open to specific institutions through a trusted access mechanism.

The new model is now fully available in the API and Enterprise Editions. Subscribed users can enjoy a free trial until June 22nd, after which, starting June 23rd, they will need to use consumption points for payment until regular subscriptions are reinstated when future computing power is sufficient.

举报 Correction/Report
RelatedRecommended
The Most Pro-Regulation AI Company Faces Regulatory Crackdown
Anthropic has opened up its Mythos-level Model Fable 5 to ordinary users, implementing permissioned access to restrict high-risk capabilities.
Mythos has now acquired stand-alone attack capability, and the risk is transitioning from containable to unpredictable
Anthropic's flagship model achieved near-total penetration in an NSA Red Team test, prompting the government to issue a ban on the same day.
Correction/Report
Submit
Add Library
Visible to myself only
Public
Save
Choose Library
Add Library
Cancel
Finish