Altman warns of possible disruptions as GPT-5.6 Sol demand strains OpenAI

OpenAI CEO Sam Altman warned users on Monday that the company's newly released GPT-5.6 Sol model may face service disruptions as demand overwhelms its inference capacity, just days after the flagship model became broadly available to the public.

"5.6 sol growth is insane," Altman wrote on X on July 14. "The inference team has done heroic work to be able to support demand. We are going to move mountains to continue to scale, but it is possible there are some hiccups soon."mlq+1

A Model Built for Agentic Work

GPT-5.6 Sol launched publicly on July 9 after a roughly two-week government-gated safety review that limited initial access to about 20 vetted partner organizations. The model is the most capable in the GPT-5.6 family, which also includes the mid-tier Terra and the lightweight Luna.axios+2

Altman told CNBC that Sol is 54% more token efficient on agentic coding tasks, a metric he framed in terms of enterprise cost consciousness. "Every enterprise now is thinking about spend and the value they're getting in exchange for AI," he said. The model introduces a "max" reasoning mode for deep problem-solving and an "ultra" mode that delegates subtasks to coordinated subagents.openai+2

Scaling Under Pressure

OpenAI has already taken steps to manage the surge. The company removed a five-hour rate limit and reset usage caps for Plus, Pro, Business, and Enterprise subscribers. It is also running Sol on Cerebras hardware at speeds of up to 750 tokens per second for enterprise customers requiring real-time inference.mlq+1

Codex and ChatGPT Work, the new productivity agent launched alongside GPT-5.6, have reached 8 million active users, up from 5 million previously, according to MLQ AI. The model is priced at $5 per million input tokens and $30 per million output tokens through the API.openai+2

From Government Gate to Open Floodgates

The broad release followed what Altman described as a "collaborative back and forth" with the Trump administration, which had asked OpenAI to delay the rollout over national security concerns related to the model's cybersecurity capabilities. TechCrunch reported in June that OpenAI called the restriction a "short-term step" and said such a process "shouldn't be the norm".reuters+2

The capacity warning underscores a recurring tension in frontier AI deployment: building models powerful enough to drive adoption while maintaining the infrastructure to serve it. For developers and enterprises relying on the OpenAI API, Altman's post signals the possibility of degraded performance or temporary service interruptions in the near term.