When it comes to AI, efficiency, accuracy, and pricing are top considerations for power users. With Cloud Ops 5, Anthropic achieved impressive results. On the performance and accuracy fronts, however, access to the model came at a premium. With the company's latest model released today, that's not a concern.

Bushra Claude announced Opus 5.5. Today, and the new model promises not only better performance and accuracy, but also a nearly 40% reduction in running costs compared to the Opus 5.

According to the company, the new model is the strongest performing model it has tested till date. Of course, with that comes concerns, which Anthropic says it's addressing with new considerations.

For starters, the new model communicates better. Anthropic says this was a concern for Opus 5 users.

According to early testers, Opus 5.5 communicates more naturally, writing is clearer and easier to follow. It's easy to analyze messages at a glance, which testers say is helpful during long work sessions. "Verbose, hard-to-follow output frontier models have been my biggest frustration, and Claude Opus 5.5 fixes that. It writes like a good companion, and follows our writing rules," wrote a Ramp software engineer, while a principal engineer at a Chicago trading company wrote, "Its writing is easier to follow and more cooperative than Opus5."

When it comes to performance, Opus 5.5 leads Fable 5.1, Opus 5, GPT-6 Astra, and GPT-5.6 Sol in benchmarks such as agentive coding, cognitive tasks, multidisciplinary reasoning, computer usage, and visual chart recognition. In benchmarks such as business workflow and agent scientific research, the new model is second only to the GPT-6 Astra, while producing output 30% faster than the Opus 5.

Of course, benchmarks only tell part of the story, so Anthropic came up with real-world receipts. According to the company, an initial tester was able to audit and fix a 200,000-line codebase in less than three hours. Opus 5 took 20 hours and 2.5x more tokens to complete the same task. Similarly, in a test that tasked an AI model with translating software from C to Rust, Opus 5.5 took 9.5 hours. Fable 5.1 took 12 hours to complete the same task.

Anthropic also shared examples of similar efficiency and intelligence gains from early testers at GitHub, Clio, Lovable, Quantum, Spotify, and more. GitHub's tester found that the model solved more terminal tasks in less than half the steps than Opus 5, while Spotify's tester hailed the model for being able to complete tasks cheaper and faster than previous models.

Elsewhere, Anthropic mentions the model's local hierarchy, which screens each action before it runs. This model also reportedly has a stronger defense than its predecessors. It reportedly matches or beats Opus 5 for reducing instant injection attacks in every setting, including coding, tool use, computer use, and web browsing. "On a benchmark run by AI security firm Gray Swan, the Opus 5.5 ties the Fable 5.1 for the lowest instant injection success rate of any model tested," the company added.

Perhaps the biggest attraction here is that Anthropic isn't charging you much for all the improvements. The API cost has dropped from $5 on Opus 5 to $4 on Opus 5.5 per million input tokens. Likewise, pricing has dropped from $25 to $20 per million output tokens. A cash-read costs $0.20 per million tokens, down from $0.50, while a cash-write costs $5 per million tokens, down from $6.25. Overall, with token performance improvements, the cost of running Opus 5.5 will be about 40% lower than Opus 5 on typical workloads, according to the company.

More importantly, you don't actually need to be a developer or use the APIs to access Cloud Ops 5.5. It's also available to regular users, with a five-hour usage limit on Pro, Max, Team, and Set-based Enterprise plans. Find out more, including detailed early tester testimonials, at Anthropic's blog post, linked above.