OpenAI launches GPT-6.1 Terra for ChatGPT Work and Codex
- Published
- Updated
OpenAI said GPT-6.1 Terra is rolling out in ChatGPT Work and Codex and will also be available through the API. The company also attached a 1,050,000-token context window to the model and said its API pricing is 50% below GPT-5.6 Terra, while its error rate fell sharply.
Key claims & sources
What it means
OpenAI is positioning GPT-6.1 Terra as part of its main product stack, with access in ChatGPT Work, Codex, and the API. That places the model in front of both business users and developers rather than keeping it in a separate release channel. The size of the context window also signals that OpenAI is aiming at long prompts and extended sessions, which matters for workflows that depend on large inputs and sustained memory across tasks.
Users building with OpenAI’s API now have a new mid-tier option that the company says is cheaper than GPT-5.6 Terra. If teams adopt it, they may rework costs and model selection inside products that rely on OpenAI services. The practical question is how far a lower-priced model can substitute for older releases without changing the behavior of apps that already depend on them.
The release reflects a wider push to turn general-purpose models into tools that can act inside software, not just answer questions. That raises the trade-off between more capable automation and tighter dependence on a single vendor’s model stack, pricing, and deployment choices. It also keeps attention on whether frontier AI products are being optimized for broad utility, lower cost, or both at once.
What remains open is how the model performs in live use compared with OpenAI’s older systems and rival offerings, beyond the vendor-reported benchmark and error-rate claims. It is also still unclear how developers will adapt once the API is in wider use and whether the new pricing sticks over time. Confirmation would come from independent testing, developer adoption patterns, and more detail on where ChatGPT Work and Codex users can access the model.
Market reaction i
Gold Park barely moved in the 4 hours after the first headline, after a flat 2 hours before.
48 prices, one every 30 minutes
Coverage timeline
Every piece in order, with the time its news medium declares — if it is not believable we use ours and say so — and what it added to what was already known at that point. Arriving first is not the same as adding something.
Decrypt
context