Anthropic launches Claude Opus 5.5, rivals Fable at 60% lower cost.

Anthropic announced on Tuesday that its new flagship model, Claude Opus 5.5, can match the performance of the high‑end Fable 5.1 system while costing roughly 60 percent less, though OpenAI’s upcoming GPT‑6 Astra still beats both on two widely cited benchmark tests. The rollout comes as the AI industry races to offer more capable models at lower prices for enterprise users.

The price gap matters because large‑language‑model services have become a major line item for tech firms, cloud providers, and developers. Anthropic’s earlier Claude 3 series was praised for safety but carried a premium price that limited adoption in cost‑sensitive sectors. Fable, a model from a boutique AI startup, earned a reputation for near‑state‑of‑the‑art accuracy but required expensive compute, making it a niche choice for only the richest customers. As AI workloads expand into customer support, data analysis, and content creation, companies are looking for cheaper alternatives that do not sacrifice quality. Anthropic’s push follows a 2024 partnership with Amazon Web Services that gave it access to cheaper hardware, and a 2025 $2 billion funding round that earmarked resources for model efficiency.

In internal tests, Opus 5.5 achieved the same scores as Fable 5.1 on the MMLU and HumanEval benchmarks while using 40 percent less token compute, translating into a roughly 60 percent reduction in per‑token cost. On the more recent BIG‑Bench and GSM‑8K tests, OpenAI’s GPT‑6 Astra posted higher accuracy and faster inference, keeping it ahead on those two metrics. Analysts at Bloomberg Intelligence said the move “tightens the competitive field” and could force OpenAI to adjust pricing. Early adopters such as a European fintech firm reported that Opus 5.5 cut their monthly AI spend by about $12,000 without noticeable drops in output quality.

The launch sets the stage for a rapid next round of model releases. Anthropic has signaled that a 6.0 version is in development, aiming to close the remaining gaps with GPT‑6. Meanwhile, OpenAI plans to roll out Astra to its enterprise tier later this year, likely with a pricing structure that reflects the new cost pressures. Industry watchers expect a short‑term price war that could lower barriers for smaller businesses to adopt advanced language models, while also accelerating research into more efficient architectures. The outcome will shape how quickly AI moves from flagship labs into everyday corporate workflows.