Add Android Authority on Google:
TL;DR
- Anthropic’s new Claude Opus 5.5 guarantees improved efficiency, accuracy, communication, and stronger security safeguards.
- The mannequin generates responses over 30% sooner and prices roughly 40% much less to run than Opus 5.
- Anthropic is rising utilization limits for Professional, Max, Crew, and Enterprise subscribers.
In the case of AI, efficiency, accuracy, and pricing are the most important concerns for energy customers. With Claude Opus 5, Anthropic achieved spectacular outcomes on the efficiency and accuracy fronts, however entry to the mannequin got here at a premium. With the corporate’s newest mannequin launched at the moment, that’s much less of a priority.
Anthropic introduced Claude Opus 5.5 at the moment, and the brand new mannequin guarantees not solely higher efficiency and accuracy, but additionally a roughly 40% discount in working prices in comparison with Opus 5.
In response to the corporate, the brand new mannequin is the strongest-performing mannequin it has examined up to now. After all, with that come issues, which Anthropic says it’s tackling with new safeguards.
For starters, the brand new mannequin communicates higher. Anthropic says this was a priority for customers with Opus 5.
In response to early testers, Opus 5.5 communicates extra naturally, with its writing being clearer and simpler to observe. Messages are a lot simpler to parse at a look, which, based on testers, is useful throughout lengthy working classes. “Verbose, hard-to-follow output has been my largest frustration with frontier fashions, and Claude Opus 5.5 fixes it. It writes like a superb colleague, and follows our writing guidelines,” wrote a Ramp software program engineer, whereas a principal engineer on the Chicago Buying and selling Firm wrote, “Its writing is simple to observe and extra coherent than Opus 5’s.”
In the case of efficiency, Opus 5.5 leads fashions like Fable 5.1, Opus 5, GPT-6 Astra, and GPT-5.6 Sol in benchmarks like agentic coding, information work, multidisciplinary reasoning, pc use, and visible chart recognition. In benchmarks like enterprise workflows and agentic scientific analysis, the brand new mannequin ranked second, solely behind GPT-6 Astra, all whereas producing output 30% sooner than Opus 5.
After all, benchmarks solely inform a part of the story, so Anthropic introduced real-world receipts. In response to the corporate, an early tester was in a position to audit and repair a 200,000-line codebase in underneath three hours. Opus 5 took over 20 hours and a couple of.5x as many tokens to finish the identical process. Equally, in a check that tasked the AI mannequin with translating software program from C into Rust, Opus 5.5 took 9.5 hours. Fable 5.1 took 12 hours to finish the identical process.
Anthropic additionally shared examples of comparable effectivity and intelligence beneficial properties from early testers at GitHub, Clio, Lovable, Quantum, Spotify, and extra. The tester at GitHub discovered that the mannequin solved extra terminal duties than Opus 5 in lower than half the steps, whereas the tester at Spotify hyped the mannequin for with the ability to full duties for cheaper and sooner than earlier fashions.
Elsewhere, Anthropic touted the mannequin’s native classifier, which primarily screens each motion earlier than it runs. The mannequin additionally reportedly has stronger defenses than predecessors. It reportedly matches or beats Opus 5 in mitigating immediate injection assaults in each setting, together with coding, device use, pc use, and net shopping. “On a benchmark run by the AI safety agency Grey Swan, Opus 5.5 ties Fable 5.1 for the bottom immediate injection success charge of any mannequin examined,” the corporate added.
Maybe the most important attraction right here is that Anthropic isn’t charging you extra for all of the enhancements. API pricing has dropped from $5 on Opus 5 to $4 on Opus 5.5 per million enter tokens. Equally, pricing has dropped from $25 to $20 per million output tokens. Cache reads value $0.20 per million tokens, down from $0.50, whereas cache writes value $5 per million tokens, down from $6.25. In response to the corporate, total, with enhancements in token effectivity, working Opus 5.5 will value roughly 40% lower than Opus 5 on typical workloads.
Extra importantly, you don’t actually have to be a developer or use APIs to entry Claude Opus 5.5. It’s obtainable to common customers too, with five-hour utilization limits on Professional, Max, Crew, and seat-based Enterprise plans. Discover extra info, together with detailed early tester testimonials, on Anthropic’s weblog publish, linked up high.
Thanks for being a part of our group. Learn our Remark Coverage earlier than posting.
