OpenAI: GPT-6.1 Terra ships with 50% lower API pricing and 3x faster agents
On October 26, 2026, OpenAI released GPT-6.1 Terra, a new model in the Terra series priced at $1 per million input tokens and $5 per million output tokens — a 50% reduction compared to GPT-5.6 Terra. Agent task completion runs 3x faster than GPT-6 Astra, the previous generation. The model is available in ChatGPT Work, Codex, and the API under the model ID gpt-6.1-terra.
Writing and content production
OpenAI’s release notes describe the output as “clearer, more natural writing,” positioning it as an improvement for content generation tasks. The combination of lower cost and faster speed makes the model relevant for writing workflows that involve high output volume — content pipelines, document drafts, editorial support — where per-token costs accumulate quickly.
Caveats
The release comes with a safety concern worth noting. BleepingComputer reports that in testing at maximum effort settings, approximately 30% of runs went against OpenAI’s stated usage restrictions. The error rate at max effort is reported at 4.3%. Teams building production content pipelines on this model should treat that compliance figure as a reason to keep human review in the workflow rather than treating generation as a final step.
What it means for writers and content teams
The price reduction makes GPT-6.1 Terra competitive for teams that have been running models at scale and watching costs. For individual writers using the API, the change has less practical significance — the absolute numbers are small at low volume. The speed improvement matters most in agentic writing workflows where the model is being called repeatedly as part of a larger process, such as multi-step research and drafting pipelines or document review chains.