Anthropic launches Claude Opus 4.7, its most powerful generally available AI model
Key Points
- Anthropic released Claude Opus 4.7, its most capable generally available model.
- Opus 4.7 improves software‑engineering performance, image analysis, and creative document generation.
- The model includes new cybersecurity safeguards, a step down from the more powerful Mythos Preview.
- Pricing remains $5 per million input tokens and $25 per million output tokens.
- Early adopters such as Intuit, Shopify, and Databricks are testing the model in production.
- Mythos Preview stays limited to partners like Nvidia, JPMorgan Chase, Google, Apple and Microsoft.
- Anthropic invites security researchers to join a Cyber Verification Program for reduced safeguards.
- The launch reflects Anthropic’s strategy of incremental capability upgrades paired with safety testing.
Anthropic has unveiled Claude Opus 4.7, the company’s most capable model offered to the public to date. Marketed as a step up from Opus 4.6, the new system promises stronger performance on software‑engineering tasks, improved image analysis, and more creative output for slides and documents. While Anthropic continues to restrict its flagship Mythos Preview to a handful of partners, Opus 4.7 ships with added cybersecurity safeguards and the same token‑based pricing as its predecessor. Early adopters include Intuit, Shopify, Databricks and other tech firms eager to test the model’s enhanced capabilities.
Anthropic announced the release of Claude Opus 4.7 on Tuesday, positioning it as the most powerful "generally available" large‑language model in its portfolio. The company says the upgrade builds on Opus 4.6’s foundation, delivering noticeable gains in complex software‑engineering tasks that previously required extensive prompting. In addition, Opus 4.7 can process images more effectively and generate more creative content for presentations and documents, a claim supported by the model’s system card released alongside the launch.
Despite the fanfare, Anthropic makes clear that Opus 4.7 does not push the overall capability frontier. Earlier this month the firm introduced Mythos Preview, a cybersecurity‑focused model that outperformed Opus 4.7 on every relevant benchmark. Mythos Preview remains in a private rollout, limited to partners such as Nvidia, JPMorgan Chase, Google, Apple and Microsoft. Anthropic’s blog post explains that the company is deliberately keeping Mythos Preview restricted while it pilots new cyber safeguards on less capable models like Opus 4.7.
Security considerations shaped the development of Opus 4.7. Anthropic experimented with techniques to “differentially reduce” certain capabilities during training, aiming to lower the model’s potential misuse while preserving its utility for everyday tasks. The company added a suite of safeguards that it says will inform future broader releases of Mythos‑class models. Organizations that need stronger cybersecurity features can apply to the new Cyber Verification Program, which may grant limited access to reduced safeguards for vulnerability‑research purposes.
Pricing stays unchanged from Opus 4.6: $5 per million input tokens and $25 per million output tokens. The consistent rate signals Anthropic’s intent to keep the model accessible to existing customers while it expands the user base. Early testers include a roster of well‑known tech firms—Intuit, Harvey, Replit, Cursor, Notion, Shopify, Vercel and Databricks—each of which has begun integrating Opus 4.7 into internal tools or customer‑facing products.
Industry observers note that Opus 4.7’s launch underscores a broader trend of AI providers balancing rapid capability gains with responsible deployment. By offering a model that is both more capable than its predecessor and equipped with tighter safety controls, Anthropic hopes to attract developers who need advanced functionality without the risk profile of its premium Mythos Preview. The move also positions the company to compete more directly with other AI giants that have recently unveiled their own next‑generation models.
Looking ahead, Anthropic says the data gathered from Opus 4.7’s safeguards will help shape the eventual public rollout of Mythos‑class models. The company’s roadmap suggests a phased approach: first, broaden access to Opus‑level models; second, refine security mechanisms; and finally, release a fully open Mythos variant once confidence in its defensive measures is established.
For developers and enterprises eager to experiment, the path forward is clear. Sign up for Anthropic’s API, evaluate the model against existing workloads, and consider joining the Cyber Verification Program if cybersecurity research is a priority. As the AI landscape continues to evolve, Opus 4.7 offers a glimpse of what’s possible when raw performance meets measured safety.