Z.ai ups ante in open-source LLMs with GLM-5.1

Z.ai launched GLM-5.1, its flagship open-source model that claims to outperform OpenAI's GPT-5.4, Anthropic's Opus 4.6 and Google's Gemini 3.1 Pro on SWE Bench Pro benchmarks for software engineering. Z.ai's GML-5.1 is competitive in other areas.

The company's next-gen model is able to perform agentic tasks over longer horizons, breaks complex problems down and improves reasoning over its GLM-5 predecessor.

Z.ai's latest efforts highlight how open-source models are becoming a battleground with US players just creeping back into a race that is dominated by Chinese models Qwen, Kimi and DeepSeek. Google launched Gemma 4 last week and Nvidia has been pushing its Nemotron models.

In a blog post, Z.ai said:

"GLM-5.1 is built to stay effective on agentic tasks over much longer horizons. We've found that the model handles ambiguous problems with better judgment and stays productive over longer sessions. It breaks complex problems down, runs experiments, reads results, and identifies blockers with real precision. By revisiting its reasoning and revising its strategy through repeated iteration, GLM-5.1 sustains optimization over hundreds of rounds and thousands of tool calls. The longer it runs, the better the result."

The consensus is that open source LLMs are roughly 6 months behind proprietary LLMs, but that gap could close. US enterprises may be restricted from using Chinese open source LLMs but can benefit if US giants become more competitive. For now, the leaderboards on Hugging Face and Arena have Z.ai's GLM 5.1 at the top following a short run by Gemma 4.

GLM-5.1 is released as open source under the MIT License. GLM-5.1 is also available on the developer platforms api.z.ai and BigModel.cn and compatible with Claude Code and OpenClaw.

• Google launches Gemma 4 open-source LLM family
• Nvidia Nemotron: Much needed open-source model champion in US
• Why enterprise AI leaders need to bank on open-source LLMs
• Nvidia launches Nemotron 3 open models to enable multi-agent systems
• Nvidia GTC 2026: Nvidia launches NemoClaw, eyes to pair with DGX Spark, DGX Station
• AI Forum 2026: "There are claws everywhere now"

Z.ai, which was formerly Zhipu AI, went public on the Hong Kong Stock Exchange in January. The company said it has developed its AI products including AutoGLM, an autonomous AI agent, AutoClaw and Zread.ai. It primarily monetizes through its Bigmodel model-as-a-service offering and APIs.

The company reported fiscal 2025 revenue growth of 131%. Based on exchange rates on March 31, the date of Z.ai's report, the company reported a net loss of $682.7 million on revenue of $104.8 million.

By segment, Z.ai's enterprise LLM revenue for fiscal 2025 was $52.9 million followed by open platform and API revenue of $27.5 million and enterprise AI agent revenue of $24 million.

Z.ai said it is focused on TAC (Token Architecture Capability), which aims to build complex agent systems within a given budget, LLM-OS where models act as a core engine to schedule compute, and a global factory for tokens leveraging China's industrial knowhow in energy and co-design.

Note: Gemini was used for translation and currency. The conversions use a rate of 1 RMB = $0.1447 USD and 1 HKD = $0.1276 USD.