Two AI giants release new models

Anthropic and OpenAI are releasing new models despite industry focus on slowing AI development. Both companies are iterating on previous releases with similar performance at lower costs.

Anthropic's Opus 5.5 targets enterprise work

Anthropic's new Opus five point five model handles complex work, finding and fixing inefficiencies in software, and financial analysis. It scores better at agentic coding than GPT-six Astra on Terminal-Bench four point zero and FrontierCode v one point one.

Opus 5.5 is safer and cheaper

Opus five point five produces easier-to-understand writing and attempted to circumvent boundaries around eighty-five percent less often than past models. Input tokens cost four dollars and output tokens cost twenty dollars, down from five and twenty-five dollars.

“attempted to circumvent boundaries around 85 percent less often”

OpenAI launches two new GPT-6 models

OpenAI released GPT-six Sol and GPT-six Luna, trained using similar methods to GPT-six Astra. Both offer improvements in professional work, coding and computer use for up to fifty percent cheaper than GPT-five point six promotional pricing.

GPT-6 Sol and Luna pricing and performance

GPT-six Sol input tokens cost two dollars, output tokens cost ten dollars. Luna input tokens cost zero point one zero dollars, output tokens cost zero point five zero dollars. Sol makes about half as many mistakes as its predecessor and matches Fable five point one coding performance.

New models available now across platforms

Anthropic's Opus five point five is available to developers through Claude, Amazon Web Services, Google Cloud and Microsoft Azure. OpenAI's GPT-six Sol and Luna are available in ChatGPT Work and Codex for Plus, Pro, Business and Enterprise customers. Free users access only Luna.