OpenAI has released its new mid-tier model, GPT-6.1 Terra, which debuted at number 14 on the Arena.ai Code Arena WebDev leaderboard with a score of 1,671 points. This result is 151 points higher than the performance of its predecessor, GPT-5.6 Terra, which currently holds the 56th position.
The new model offers improved cost efficiency, with a median task cost of $0.40 compared to $0.56 for the previous version, representing a 28% reduction. OpenAI has also lowered API pricing by 50% relative to the predecessor, setting the rate at $1 per million input tokens and $5 per million output tokens.
In performance testing on Terminal-Bench 4.0, GPT-6.1 Terra successfully completed 50.3% of tasks, significantly outperforming the 21.5% success rate of GPT-5.6 Terra. Additionally, the model completes tasks more than three times faster on average than the GPT-6 Astra and GPT-6.1 Sol models.
OpenAI positions GPT-6.1 Terra as a solution for visual reasoning and routine computer tasks. The model is currently available through ChatGPT Work, the Codex platform, and via the API under the identifier gpt-6.1-terra.
