New Model Delivers Fable-Level Performance at Reduced Rates

Anthropic introduced Claude Opus 5.5 as the first member of its Claude 5.5 model family. The company states the model reaches performance levels close to its top-tier Claude Fable 5.1 on most tasks while charging substantially less. Input tokens are priced at $4 per million and output tokens at $20 per million. That represents roughly a 40 percent reduction compared with the prior Opus 5 rates of $5 and $25. Against Fable 5.1, which costs $10 input and $50 output, the new model runs about 60 percent cheaper.

Cached reads for Opus 5.5 are set at $0.20 per million tokens, a 60 percent decline from earlier Opus pricing. Anthropic also reports that response generation is more than 30 percent faster than the previous version. The model is designed for coding, research, and professional knowledge work.

Benchmark Results and Practical Coding Gains

On Terminal-Bench 4.0, which measures an agent’s ability to complete complex professional tasks in a command-line setting, Opus 5.5 scored 66.4 percent. That exceeds Opus 5’s 52.3 percent, Fable 5.1’s 55.8 percent, and GPT-6 Astra’s 57.9 percent. On FrontierCode, which evaluates whether code changes would be accepted in a real engineering pipeline, the new model reached 54.4 percent versus Fable 5.1’s 50.3 percent. GDPval-AA v2.1, an Elo-based ranking of professional work across 44 occupations, gave Opus 5.5 a score of 1,846, ahead of Fable 5.1’s 1,735 and Opus 5’s 1,708. CursorBench 4.0 registered 57.8 percent.

The model trails GPT-6 Astra on two other tests: AutomationBench (40.0 percent versus 41.4 percent) and Terminal-Bench-Science 0.1 (58.7 percent versus 64.6 percent). Anthropic notes that small differences on these metrics matter less at current capability levels and that real-world outcomes can vary.

In practical trials, one tester completed a 680,000-line code migration in under a day. Another finished a 200,000-line code audit and repair in less than three hours, work that previously required more than 20 hours with Opus 5.

Competitive Pricing Moves and Broader Industry Pressure

The same day Anthropic released Opus 5.5, OpenAI introduced GPT-6 Sol and GPT-6 Luna with API pricing about 50 percent below earlier ChatGPT-5.6 levels. Both companies face growing competition from lower-cost open-weight models developed by firms including Alibaba, Moonshot AI, and DeepSeek. Industry data indicate these providers are capturing share from U.S. laboratories, prompting a shift toward cost competition alongside raw capability.

Anthropic’s release is the first model shipped since CEO Dario Amodei publicly argued that AI capability gains are outpacing safety testing. The company states Opus 5.5 underwent additional external evaluation and carries the same cybersecurity and biology safeguards applied to Fable 5.1. Most cybersecurity tasks are automatically routed to an earlier model. OpenAI CEO Sam Altman and Elon Musk have expressed support for greater caution, while President Donald Trump stated at the United Nations that the United States would not restrain AI growth, though law enforcement could intervene if issues arise.

IPO Preparations and Company Trajectory

The product launch arrives as Anthropic prepares for a potential public listing in November. Reports indicate the company may seek a valuation near $2 trillion and could raise up to $100 billion, though final terms have not been confirmed. Earlier expectations of an October listing were adjusted to allow stronger third-quarter results to be shown.

Anthropic’s valuation has risen rapidly: approximately $61.5 billion in early 2025, $183 billion in September 2025, $380 billion after a Series G round in February, and $965 billion after a $65 billion Series H financing in May that included Amazon, Micron, Samsung, and SK Hynix. The firm’s annualized revenue run rate reached about $65 billion by July, up from roughly $9 billion at the end of the prior year.

Claude Opus 5.5 is available immediately on the Claude platform as well as Amazon Web Services, Google Cloud, and Microsoft Azure. Anthropic plans to introduce Claude Sonnet 5.5 and Claude Haiku 5.5 in the coming weeks with corresponding price reductions across the lineup.