OpenAI Releases GPT-6.1 Terra Model

en.Wedoany.com Reported - On October 26, the American artificial intelligence company OpenAI officially released the GPT-6.1 Terra model, with key upgrades in visual reasoning, computer operation, and natural language processing capabilities. The new model targets technical drawing analysis, software operation, and enterprise automation tasks. Standard API input pricing has been reduced to $1 per million tokens, with output pricing at $5 per million tokens, further lowering usage costs compared with the previous-generation model.

The new model can recognize application interfaces and execute multi-step operations, and supports drafting business proposals, interpreting engineering technical drawings, filling out electronic forms, and assisting with video editing. In the professional task benchmark Agents' Last Exam, GPT-6.1 Terra at medium reasoning effort scored 15 percentage points higher than GPT-5.6 Terra, with task processing costs approximately one-third of the latter. Compared with GPT-6 Astra and GPT-6.1 Sol, it completes comparable test tasks more than three times faster.

Visual reasoning performance has also been upgraded. In the MMMU-Pro test, GPT-6.1 Terra outperforms the previous-generation model at all reasoning effort levels. OpenAI's internal factual accuracy tests show that the error rate at low reasoning effort dropped from 53.6% to 16.7%. In the Terminal-Bench 4.0 software engineering test, the new model achieved a task completion rate of 50.3% at the highest reasoning effort, compared with 21.5% for the previous-generation model.

GPT-6.1 Terra supports a context window of approximately 1.05 million tokens, with a maximum output length of 128,000 tokens. Developers can access it via the OpenAI API under the model identifier gpt-6.1-terra. The model is also available to eligible ChatGPT Work and Codex users and was officially launched on the Amazon Web Services Amazon Bedrock platform on October 26.