most agent labs are shy about acknowledging chinese model use because they need to sell to gov/defense cog team did the hard part to productionize:…

Cognition released SWE-1.7, a model trained using reinforcement learning that achieves performance parity with frontier models at a significantly lower cost. The deployment infrastructure supports throughput of 1,000 tokens per second. Technical developments include a multilingual propaganda and censorship evaluation framework used to refine post-training performance.