- Anthropic launched Claude Opus 4.5, its new flagship AI model focused on coding performance and price reduction.
- Claude Opus 4.5 outscored competitors, including OpenAI‘s GPT-5.1-Codex-Max and Google’s Gemini 3 Pro, on the SWE-bench Verified software engineering benchmark.
- The model supports a 200,000-token context window and can output up to 64,000 tokens, with a knowledge cutoff in March 2025.
- Anthropic offers a three-model lineup—Opus, Sonnet, and Haiku—catering to various development needs from complex tasks to speed and efficiency.
- Pricing for Opus 4.5 has been reduced by 67%, positioned competitively among other major AI platform costs.
Anthropic released Claude Opus 4.5 on Monday, completing its three-model family and marking the company’s third major launch in two months. The new model claims the top spot in coding benchmarks while offering significantly lower prices.
This launch follows a rapid sequence starting with Claude Sonnet 4.5 in late September and Claude Haiku 4.5 in October. Now, Anthropic provides three specialized models: Opus for complex production work, Sonnet for general tasks, and Haiku for fast, simple logic tasks.
Claude Opus 4.5 scored 80.9% on the SWE-bench Verified test that evaluates real-world software engineering challenges. This surpasses OpenAI’s GPT-5.1-Codex-Max at 77.9% and Google’s Gemini 3 Pro at 76.2%. Internally, Anthropic reported that Opus outperformed all human test-takers on a two-hour performance engineering exam measuring judgment under pressure.
The launch moves swiftly amid AI competition, with Google introducing Gemini 3 Pro in November and OpenAI launching GPT-5.1-Codex-Max the following day. Anthropic priced Opus 4.5 at $5 per million input tokens and $25 per million output tokens, a 67% price reduction from the previous Opus version. This positions it between OpenAI’s GPT-5.1 pricing ($1.25/$10) and its own older Opus 4.1 model ($15/$75), though still higher than Gemini 3 Pro’s $2/$12 rates.
Alibaba also competes with its Qwen series, including Qwen2.5-Max and Qwen3-Max, which focus on diverse tasks and rank among the top AI models globally, though their presence is stronger in Eastern markets. These efforts align with China’s goal for AI self-reliance amid U.S. chip export limits.
The Opus 4.5 model incorporates Anthropic’s “hybrid reasoning” architecture, a single system trained for both direct inference and chain-of-thought processes. It supports a 200,000-token context window and can generate up to 64,000 tokens in output. Its data knowledge cutoff is March 2025, ahead of Sonnet 4.5’s January cutoff.
Developer Simon Willison tested Opus 4.5 on a project involving 20 commits across 39 files, making substantial code changes. Willison described it as an excellent model but noted the productivity gains over Sonnet 4.5 were not dramatically clear. AI platform CEO Theo Browne called Claude Opus 4.5 “insane” and “definitely the best coding model ever made” in a video review.
Recently, Microsoft and NVIDIA announced multi-billion-dollar investments in Anthropic, raising its valuation to around $350 billion. These investments include enhanced integration with Microsoft Azure and Nvidia infrastructure for training and deploying Claude models.
Claude Opus 4.5 is accessible immediately through Anthropic’s API, AWS Bedrock, Google Vertex AI, and Claude’s web and desktop applications.
✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.
Previous Articles:
- Thousands of Passwords Exposed via Popular JSON Formatter Tools
- Polymarket Gets CFTC Nod to Operate US Retail Exchange
- Elon Musk’s xAI Extends $15B Raise, Now Targeting $230B Value
- Viva La Libertad website taken down amid LIBRA token probe
- Standard Chartered Becomes 21Shares’ New Digital Asset Custodian
