Affiliate links on Android Authority may earn us a commission. Learn more.
When it comes to AI, performance, accuracy, and pricing are the biggest considerations for power users. With Claude Opus 5, Anthropic achieved impressive results on the performance and accuracy fronts, but access to the model came at a premium. With the company’s latest model released today, that’s less of a concern.
Anthropic announced Claude Opus 5.5 today, and the new model promises not only better performance and accuracy, but also a roughly 40% reduction in running costs compared to Opus 5.
According to the company, the new model is the strongest-performing model it has tested to date. Of course, with that come concerns, which Anthropic says it is tackling with new safeguards.
For starters, the new model communicates better. Anthropic says this was a concern for users with Opus 5.
According to early testers, Opus 5.5 communicates more naturally, with its writing being clearer and easier to follow. Messages are much easier to parse at a glance, which, according to testers, is helpful during long working sessions. “Verbose, hard-to-follow output has been my biggest frustration with frontier models, and Claude Opus 5.5 fixes it. It writes like a good colleague, and follows our writing rules,” wrote a Ramp software engineer, while a principal engineer at the Chicago Trading Company wrote, “Its writing is easy to follow and more coherent than Opus 5’s.”
When it comes to performance, Opus 5.5 leads models like Fable 5.1, Opus 5, GPT-6 Astra, and GPT-5.6 Sol in benchmarks like agentic coding, knowledge work, multidisciplinary reasoning, computer use, and visual chart recognition. In benchmarks like business workflows and agentic scientific research, the new model ranked second, only behind GPT-6 Astra, all while generating output 30% faster than Opus 5.
Of course, benchmarks only tell part of the story, so Anthropic brought real-world receipts. According to the company, an early tester was able to audit and fix a 200,000-line codebase in under three hours. Opus 5 took over 20 hours and 2.5x as many tokens to complete the same task. Similarly, in a test that tasked the AI model with translating software from C into 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. The tester at GitHub found that the model solved more terminal tasks than Opus 5 in less than half the steps, while the tester at Spotify hyped the model for being able to complete tasks for cheaper and faster than previous models.
Elsewhere, Anthropic touted the model’s native classifier, which essentially screens every action before it runs. The model also reportedly has stronger defenses than predecessors. It reportedly matches or beats Opus 5 in mitigating prompt injection attacks in every setting, including coding, tool use, computer use, and web browsing. “On a benchmark run by the AI security firm Gray Swan, Opus 5.5 ties Fable 5.1 for the lowest prompt injection success rate of any model tested,” the company added.
Perhaps the biggest attraction here is that Anthropic isn’t charging you more for all the improvements. API pricing has dropped from $5 on Opus 5 to $4 on Opus 5.5 per million input tokens. Similarly, pricing has dropped from $25 to $20 per million output tokens. Cache reads cost $0.20 per million tokens, down from $0.50, while cache writes cost $5 per million tokens, down from $6.25. According to the company, overall, with improvements in token efficiency, running Opus 5.5 will cost roughly 40% less than Opus 5 on typical workloads.
More importantly, you don’t really need to be a developer or use APIs to access Claude Opus 5.5. It is available to regular users too, with five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans. Find more information, including detailed early tester testimonials, on Anthropic’s blog post, linked up top.
Thank you for being part of our community. Read our Comment Policy before posting.
