Anthropic releases Claude Opus 4.7 with new safeguards
SkimNews Take
Anthropic's public Opus release, while an advance, highlights a growing gap between cutting-edge internal AI capabilities and what is deemed safe or stable enough for broader commercial deployment.
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic released Claude Opus 4.7, its most powerful generally available model to date.
- Claude Opus 4.7 is positioned as a step up from Opus 4.6 for advanced software engineering, image analysis, instruction following, and creative slide/document generation, but does not advance the overall capability frontier compared to Mythos Preview.
- Mythos Preview remains a more powerful, cybersecurity‑focused model that is only available privately to select partners (Nvidia, JPMorgan Chase, Google, Apple, Microsoft) while Anthropic tests new safeguards on less capable models.
- Anthropic added extra cybersecurity safeguards to Opus 4.7 and will use deployment findings to work toward a broader release of Mythos‑class models.
- Cyber Verification Program invites security professionals to join and access some of the safeguards introduced in Opus 4.7 for vulnerability research.
- Early testers of Opus 4.7 include Intuit, Harvey, Replit, Cursor, Notion, Shopify, Vercel, and Databricks.
- Pricing for Opus 4.7 stays at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.6.
Why it matters: Security teams gain early access to a model with built‑in cyber safeguards and a verification program, while Anthropic gathers deployment data to pave the way for a wider Mythos‑class rollout; token pricing stays at $5 per million input and $25 per million output, and early adopters such as Intuit and Shopify can test the new capabilities.
Ask SkimNews
