Slow and steady? Anthropic and OpenAI drop new powerful models

Anthropic released Claude Opus 5.5 on Tuesday, calling it "a major step up from Opus 5" and the strongest-performing model on its automated behavioral audit, about an hour before OpenAI released GPT-6-Sol and GPT-6-Luna, cheaper models less capable than OpenAI's top-of-the-line GPT-6-Astra. Both releases came after weeks of debate over AI model performance risks and alignment concerns tied to the OpenAI/Hugging Face hacking incident, with Anthropic noting Opus 5.5 is its first model since CEO Dario Amodei called to "pace the frontier.

After weeks of debate about the potential harms from continuing to push the boundaries of AI model performance, Anthropic and OpenAI released new models Tuesday that push the boundaries of AI model performance. Anthropic's Claude Opus 5.5 "is a major step up from Opus 5," the company said in a blog post https://www.anthropic.com/claude-opus-5-5?ref=thestack.technology Tuesday morning, while also acknowledging that it is the first model Anthropic has released since CEO Dario Amodei suggested it was time to " pace the frontier. https://darioamodei.com/post/we-must-pace-the-frontier?ref=thestack.technology " About an hour later, OpenAI released https://openai.com/index/introducing-gpt-6-sol-and-luna/?ref=thestack.technology GPT-6-Sol and GPT-6-Luna, which are less capable than OpenAI's top-of-the-line GPT-6-Astra model but can be run far more cheaply, especially after a price cut. With the new releases, both companies nodded to the "alignment" issue, which has taken on international importance over the last several weeks https://www.thestack.technology/altman-amodei-musk-call-for-ai-slowdown-as-hugging-face-warns-over-closed-doors/ as the extent of the OpenAI/Hugging Face hacking incident became clear and researchers from both companies expressed concerns that there's a non-zero chance AI could cause catastrophic outcomes. "On our automated behavioral audit, the most comprehensive alignment test we run, Opus 5.5 is the strongest-performing model we’ve tested to date," Anthropic said. For OpenAI's part, it said "in our alignment evaluations, both Sol and Luna show improvements over their GPT‑5.6 counterparts, including lower rates of misleading claims about their coding work."