What happened
On September 29, 2026, OpenAI held DevDay 2026, which it called its biggest yet, with more than 20 announcements across ChatGPT, Codex and models. Three matter directly to developers.
GPT-6.1 Sol
GPT-6.1 Sol is a major upgrade to GPT-6 Sol with strong performance in agentic coding, computer use and professional work. OpenAI says it delivers near-Astra intelligence at a fifth of the standard input and output token prices.
Ultrafast
GPT-6 Astra Ultrafast became available in the API and in ChatGPT Work and Codex on Pro 500 and Enterprise plans. GPT-6.1 Sol Ultrafast was announced as "coming soon" (later reporting says it opened to all API users on October 8).
GPT-Live-1 in the API (September 10)
Weeks earlier, on September 10, OpenAI brought GPT-Live-1 to the API. It is a full-duplex voice model that listens and speaks at the same time and can delegate deeper reasoning and actions to other models and tools. It costs $0.05 per minute for the voice layer; the backend model is billed separately.
Why it matters
"Near-frontier model, much cheaper" plus "cheap real-time voice" lowers the cost of service and automation agents. The bill now depends more on architecture (which model for which step) than on the model itself.
What to do now
- Re-evaluate flows using Astra: 6.1 Sol may cut costs without meaningful loss.
- For voice projects, add GPT-Live-1 to the backend model cost to estimate the real price per minute.
- Use Ultrafast only where latency truly affects users.