According to OpenAI, GPT-6.1 Sol offers improvements compared to the earlier GPT-6 Sol. The company says the model is better at coding and debugging, understanding documents and handling tasks that require multiple steps. In one coding test, DeepSWE 1.1, GPT-6.1 Sol reportedly matched the performance of GPT-6 Astra. However, OpenAI says the new model costs around one-fifth as much per task in this benchmark.
