gpt-5.6-terra

GPT-5.6-Terra Debuts with Strong Performance Across Benchmarks

Azure OpenAI's GPT-5.6-Terra establishes its baseline performance with impressive results across multiple evaluation domains. The model demonstrates exceptional capabilities in mathematical reasoning, achieving 95.2% on GSM8K and 88.7% on MATH, positioning it among the strongest performers for quantitative tasks. Code generation shows solid competency with 87.3% on HumanEval and 84.1% on MBPP, indicating reliable programming assistance capabilities. General knowledge and reasoning metrics are robust, with 89.4% on MMLU and 86.2% on HellaSwag, while instruction following scores 88.9% on IFEval. The model handles multilingual tasks adequately at 82.3% MMMLU. Creative writing quality rates at 7.8 out of 10, suggesting strong but not exceptional performance in generative tasks. Latency metrics show 847ms time to first token and 42ms per token, indicating moderate response speeds that may impact real-time applications. With a 128K context window, the model supports substantial document processing. As a first benchmark, these results establish GPT-5.6-Terra as a well-rounded model with particular strengths in mathematics and coding, though users should monitor performance stability in future windows.