Anthropic releases its flagship AI model Opus 4.7, featuring improved software engineering capabilities and intentionally weakened cyber offense and defense functions.

```

Anthropic launched the upgraded version of its main model, Opus 4.7, just a week after restricting the release of its most advanced model, Mythos. The company is also seeking a new round of financing, with its valuation potentially doubling to $800 billion or more.

The company stated on Thursday that Opus 4.7 performs better in software engineering, with improved instruction compliance and image recognition capabilities.

However, Anthropic made it clear that this model’s overall capabilities in cybersecurity are weaker than Mythos, and the company even experimentally implemented “differentiated reduction” for its cyber offense and defense abilities during training.

This deployment strategy reflects Anthropic’s cautious trade-off between model capability and safety boundaries, and paves the way for the wide commercialization of models at Mythos’s level.

Opus 4.7: Improved Programming Ability, Limited Cyber Capabilities

Anthropic stated that the core improvement of Opus 4.7 centers on the field of software engineering, including handling high-difficulty programming tasks that previously required more human supervision. In addition, the model’s image recognition capability has also improved, allowing it to identify details in complex charts or images.

In terms of security mechanisms, Anthropic has built an automatic detection and interception system into Opus 4.7, able to identify and block requests involving prohibited or high-risk cybersecurity uses. The company said in a blog post, "We will accumulate experience from the actual deployment of these protective measures, to advance the goal of widespread release of Mythos-level models."

Notably, Anthropic actively intervened in the training phase of Opus 4.7, specifically implementing "differentiated reduction" for its cyber attack and defense abilities, a rare practice in large model development.

Concerns About Mythos Capabilities, Limited Release Becomes Established Choice

Last week, Anthropic revealed that its Mythos system possesses highly dangerous cyber attack capabilities—under user instructions, the model can identify and exploit security vulnerabilities in "every mainstream operating system and every mainstream browser."

For safety reasons, Anthropic decided to open Mythos only to select enterprise clients, specifically to help them strengthen their software systems, rather than releasing it to the general public.

Against this backdrop, the launch of Opus 4.7 not only fills the market gap that Mythos cannot serve widely, but also provides Anthropic with an opportunity to validate its security mechanisms in real deployment scenarios.

Competitive Landscape and Financing Outlook

Anthropic is currently in fierce competition with OpenAI, with both sides dedicated to deploying more advanced AI models and winning over enterprise clients for paid adoption.

In recent months, Anthropic’s AI programming products have shown strong momentum, and the consumer side continues to gain more attention, despite differences with the Pentagon on AI safety standards.

In terms of financing, Anthropic’s recent round valued the company at $380 billion. Reportedly, the company is currently soliciting bids from investors for a new round, with valuation possibly reaching about $800 billion or higher, more than double the previous round.

Risk Warning and DisclaimerThe market involves risks, and investment requires caution. This article does not constitute personal investment advice, nor does it consider the unique investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article fit their specific situation. Invest accordingly, at your own risk. ```