GPT-5.6 Sol tops Mythos in coding — so why is the US gov blocking access
openai's gpt-5.6 sol beats claude mythos in agentic coding benchmarks, but us government restrictions are currently limiting access to select partners only, delaying broader developer availability.
openai unveiled gpt-5.6 sol, a new model generation positioned against claude mythos, scoring 91.9 percent on terminal-bench 2.1 in its 'ultra' mode compared to mythos 5's 88 percent. the launch introduces a tiered naming structure—sol, terra, and luna—and claims efficiency gains in cybersecurity tasks like v8 engine exploitation. however, access is limited to select partners due to us government intervention, echoing the previous removal of anthropic's fable 5. openai criticized these restrictions, arguing they withhold essential tools from defenders and enterprises. the sol tier costs $30 per million output tokens.