OpenAI’s GPT-5.6 models ship this Thursday following a delay imposed by the U.S. government. The Department of Commerce approved the public release after the Center for AI Standards and Innovation completed additional testing on the systems. OpenAI stated the hold kept advanced tools away from developers and companies while noting that binding standards for such releases still do not exist under the current executive order.
The Sol variant beats Anthropic‘s Claude Mythos 5 on several benchmarks including TerminalBench 2.1 where it scored 88.8 percent compared to 88 percent for the competitor. On cybersecurity tasks Sol matched Mythos 5 performance while using only a third of the tokens. Sol costs $5 per million input tokens and $30 per million output tokens which is significantly cheaper than Anthropic’s Fable 5 pricing of $10 and $50 respectively. The lower token usage means users can run longer prompts without hitting limits as quickly.
* Sol Ultra reached 91.9 percent on TerminalBench 2.1
* Anthropic’s Fable 5 runs at double the cost per token
* No binding standards exist for releasing models under the executive order
