A model has been released from InternScience (Shanghai AI Research Institute) in China, with a model size of only 35b, which achieves large model performance. I haven't tested it yet, but the bench indicators alone are impressive. Some items show better results than gpt 5.5. It's a shame that I need a Vision model... (text input only)
https://internscience.github.io/Agents-A1/
▶ Original source: https://internscience.github.io/Agents-A1/
▶ Original source: https://huggingface.co/InternScience/Agents-A1
