Anthropic Completes AI Model Upgrades With Claude Opus 4.5—And Slashes Prices

Anthropic has launched Claude Opus 4.5. Mondaythe corporate’s third main launch, accomplished three mannequin households in simply two months. The brand new flagship mannequin takes prime spot in coding benchmarks at a considerably decreased worth.

This launch concludes a speedy growth that started with Claude Sonnet 4.5 in late September and continued with Claude Haiku 4.5 in October. Opus joins its sibling, and Anthropic gives builders with an entire toolkit. Opus for advanced manufacturing duties, Sonnet for mundane duties, and Haiku for pace and efficiency-related duties that require easy logic.

Claude Opus 4.5 scored 80.9%. SWE bench verifieda benchmark that checks real-world software program engineering duties. That is increased than OpenAI’s GPT-5.1-Codex-Max at 77.9% and Google’s Gemini 3 Professional at 76.2%. Anthropic stated Opus outperformed all human check takers on its inner efficiency engineering examination, a two-hour evaluation designed to evaluate judgment underneath stress.

There was a race amongst AI giants to complete the 12 months on the prime of the leaderboard. Google launched Gemini 3 Professional on November 18th, billing it as a breakthrough in multimodal inference. OpenAI responded the subsequent day with GPT-5.1-Codex-Max.

Anthropic’s reply to Opus got here just some days later, with pricing at $5 per million enter tokens and $25 per million output tokens, representing a 67% worth discount from the earlier Opus mannequin.

Alibaba’s Qwen mannequin provides one other dimension to racing. In late January, the corporate launched Qwen2.5-Max with over 20 trillion coaching tokens and claimed to outperform DeepSeek-V3 on key benchmarks. Launched in September with over 1 trillion parameters, Qwen3-Max is ranked #3 on the planet by LMArena and excels at quite a lot of duties, together with: deep analysismultimodal inference, or Japanese language workflows. Though the Qwen mannequin stays comparatively obscure in Western markets, it symbolizes China’s push towards AI independence amid U.S. chip export restrictions.

Its worth falls between OpenAI’s newest GPT-5.1 ($1.25/$10) and Anthropic’s older Opus 4.1 ($15/$75), nevertheless it’s nonetheless increased than Gemini 3 Professional’s $2/$12. This discount is indicative of market pressures as main AI labs compete not solely on capabilities but additionally on making frontier intelligence economically viable for large-scale deployment.

Claude’s newest providing continues to be dearer than lots of its Asian opponents, nevertheless it additionally affords a bit extra performance. Customers can now select between value effectivity and pure technical prowess.

sonnet 4.5, Launched on September thirtiethaffords state-of-the-art coding and agent capabilities at a modest value, and was already higher than Opus 4.1 for sure duties. An easier model of Haiku 4.5 was introduced on October fifteenth. Opus 4.5 presently sits on the prime, dealing with probably the most tough inferences and longest-running duties.

Like Sonnet and GPT-5, Claude Opus 4.5 makes use of what Anthropic calls a “hybrid inference” structure: a single mannequin skilled for each direct inference and chain-of-thought processing. It helps a 200,000 token context window and may output as much as 64,000 tokens. The information cutoff for this mannequin is March 2025, barely sooner than Sonnet’s January date.

Developer Simon Willison I examined Opus 4.5 I labored on it extensively over the weekend and used it to refactor one among my initiatives. The mannequin processed 20 commits throughout 39 recordsdata, including 2,022 rows and eradicating 1,173 different rows. “That is clearly a greater new mannequin,” Willison writes, however says that reverting again to Sonnet 4.5 would not end in a dramatic drop in productiveness.

“I am not saying the brand new mannequin is not an enchancment on Sonnet 4.5, however I am unable to say with confidence that the challenges I posed would have recognized any vital variations in performance between the 2 fashions,” he wrote.

Theo Browne, developer, YouTuber and CEO of AI platform T3 Chat, known as Claude Opus 4.5 “insane” and added: video assessment It is “arguably the most effective coding mannequin ever created.”

The aggressive panorama is changing into more and more crowded. Final week, Google’s Gemini 3 Professional dominated the headlines, scoring 1501 factors on LMArena and incomes reward from Salesforce CEO Marc Benioff, who stated he would ditch ChatGPT and mannequin it after Google. The announcement despatched Alphabet’s inventory worth up greater than 6%. reportedly OpenAI CEO Sam Altman instructed colleagues that Gemini would trigger “short-term financial headwinds.”

Microsoft and Nvidia introduced Final week, a multibillion-dollar funding in Anthropic boosted the startup’s valuation to about $350 billion. The deal contains expanded Azure integration and Nvidia-powered infrastructure for coaching and deploying cloud fashions.

Opus 4.5 is accessible instantly under. Anthropic’s APIAWS Bedrock, Google Vertex AI, Claude net and desktop apps.