Zhipu and MiniMax's latest AI models take on US giants
What's the story
Chinese start-ups Zhipu AI and MiniMax have launched their latest flagship artificial intelligence (AI) models in a bid to become leading open-source AI systems suppliers. The launches came just hours apart as both companies look to generate interest for their upcoming initial public offerings (IPOs) in Hong Kong. Both firms had cleared important regulatory hearings over the weekend.
Model launch
Zhipu AI's GLM-4.7 model takes on global giants
Beijing-based Zhipu AI, also known as Z.ai, unveiled its new GLM-4.7 model. The company claims that the new model has better coding capabilities and is closing the gap with larger US firms such as Google DeepMind and Anthropic in innovative coding agents. The firm highlighted GLM-4.7's high scores on various coding benchmarks, which surpassed those of OpenAI's GPT-5 released in August this year.
Performance
GLM-4.7 sets new benchmark
Zhipu AI's GLM-4.7 model scored just behind Google DeepMind's Gemini Pro 3 and Anthropic's Claude Opus 4.5 on various coding benchmarks. Notably, it also matched the Anthropic's older Claude Opus 4 model in breaking the 70% mark on SWE-Bench, a benchmark that tests the models on real-world software engineering problems.
New contender
MiniMax's M2.1 model also shows improvement
Hours after Zhipu AI's announcement, Shanghai-based MiniMax unveiled its M2.1 model. The new model also showed significant improvements in coding capabilities during internal tests, although it still lagged behind Anthropic's Claude Opus 4.5 on some benchmarks.