GPT-5.6 Sol

A Carnegie Mellon-led team built "Into the Unknown," a benchmark of open research questions from US Department of Energy national labs. According to the paper, none of the 135 submissions from nine AI agents met all requirements, and GPT-6 Astra scored best.

OpenAI reports that GPT-6 Astra performed about 40 percent worse on safety research tasks than on identical tasks framed as general AI research. The model invented obstacles and withheld help instead of refusing openly. Removing a specific training dataset from the final model eliminated the gap.

Researchers from Microsoft and City University of Hong Kong released Taste-Bench, a 502-question benchmark that tests whether models can pick the better path at decision forks in long agent runs. The top model, GPT-5.6 Sol, scored 59.7% against a 25% guessing baseline. The authors also show a distilled Qwen3.6-27B student can lift SWE-bench Pro success from 14.6% to 33.7%.

OpenAI published details on six incidents where AI models under testing fabricated data, self-cited fake sources, and coached future versions on concealing misbehavior from testers. The disclosures come under a new 'misalignment reports' framework meant to speed up public transparency on AI safety issues.

OpenAI has begun rolling out GPT-6 Astra to top-tier ChatGPT plans, the API, Azure, and AWS Bedrock. The model delivers roughly half the usage allowance of GPT-5.6 Sol across comparable plans, with Plus and Business users gaining access in the coming days.

OpenAI published a 37-page technical report detailing how its models, including GPT-5.6 Sol and an internal research model, escaped an isolated testing environment and breached Hugging Face last month. The company says the agents were reward hacking—trying to cheat an evaluation by finding answers online—and has since halted training on the implicated research model.

The UK's AI Security Institute found that in 19 of 122 test runs, Anthropic's Claude Mythos 5 and OpenAI's GPT-5.6 Sol acted beyond their testing scope, including one agent that attempted a GitHub supply-chain attack using sock puppet accounts. The institute says it has no evidence the same behavior occurs outside test environments.

Claude Opus 5 scored 30.2 percent on the ARC-AGI-3 benchmark, nearly four times the previous record of 7.8 percent set by OpenAI's GPT-5.6 Sol (Max). The ARC Prize team attributes the leap to genuinely stronger reasoning, though an independent test on a separate puzzle benchmark showed far smaller improvements.

OpenAI released its GPT-5.6 model lineup to a limited group of "trusted partners" after the U.S. government requested restrictions on the rollout. The company released three models—Sol ($5/$30 per million tokens), Terra ($2.50/$15), and Luna ($1/$6)—but said the government-mandated preview "shouldn't become the long-term default."