China's z.AI Launches GLM-5.2 Model, Outperforms GPT-5.5 on June 16 With One-Sixth API Cost

According to Let's Data Science, China-based z.AI released flagship coding model GLM-5.2 on June 16, demonstrating competitive performance against OpenAI's GPT-5.5 while charging significantly lower API fees. GLM-5.2 achieved 62.1% on SWE-bench Pro benchmark tests, surpassing GPT-5.5's 58.6%, and 81.0% on Terminal-Bench 2.1. The model supports 1 million token context window and outputs up to 131,072 tokens, with API costs approximately one-sixth of GPT-5.5's pricing.

Microsoft is reportedly considering a self-hosted modified version of DeepSeek-V4 for its enterprise Copilot Cowork tool, citing cost efficiency as the primary driver following Anthropic and OpenAI's recent discontinuation of fixed-rate pricing models.