Anthropic announced the release of Opus 5, positioning it as the next go‑to model for most commercial users. According to the company, Opus 5 delivers performance that nearly matches its top‑tier offering, Fable 5, but at roughly half the price. The pricing structure—$5 per million input tokens and $25 per million output tokens—represents a significant reduction from the rates associated with Fable 5, which has been a point of contention among developers.
Beyond cost, Anthropic highlights several functional improvements. Opus 5 is said to outpace earlier versions on knowledge‑work tasks, handling complex queries with greater accuracy. In coding benchmarks, the model shows a marked advantage over Opus 4.8, tackling intricate programming challenges more effectively. Researchers also benefit, as the model reportedly excels at scientific tasks such as predicting the impact of protein‑sequence variations on molecular function.
Safety and guardrails received a refresh with the new release. Anthropic claims Opus 5 exhibits the lowest rates of deceptive behavior among its models, making it harder to trick. While the model retains strong cybersecurity capabilities—identifying vulnerabilities more adeptly than its predecessor—it is deliberately kept "substantially behind" the flagship model when it comes to exploiting those vulnerabilities. The company limited training on cyber‑related tasks to avoid encouraging malicious use, and it narrowed the scope of its cyber guardrails to a specific set of activities.
Anthropic also adjusted its approach to content moderation. Earlier, the company routed certain sensitive prompts through Opus 4.8 instead of Fable 5, a move that many users criticized as overly restrictive. With Opus 5, the company says its safeguards are designed to enable beneficial applications in both cybersecurity and biology while intervening about 85 percent less often than they did for Fable 5.
Users gain additional control through a newly introduced “effort” menu. This feature lets developers specify whether they want the model to prioritize thoroughness or speed, helping conserve token usage for cost‑sensitive workloads. Opus 5 is available across all of Anthropic’s paid tiers, and there is no separate sign‑up process required.
The timing of the launch places Anthropic in direct competition with China’s Moonshot AI, which recently introduced the Kimi K3 model. Kimi K3 offers comparable performance on some tasks at $15 per million output tokens, a price point higher than Opus 5’s but still competitive. Industry observers anticipate mixed reactions: while Opus 5’s price advantage is welcomed, some may compare it unfavorably to alternative models that promise similar performance.
Overall, Anthropic expects Opus 5 to become the default choice for developers seeking a balance of capability, cost, and safety. By delivering near‑Fable performance at a reduced price and tightening its guardrails, the company aims to address long‑standing user frustrations while expanding the model’s appeal across sectors ranging from software development to biomedical research.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.