GLM-4.5: New SOTA Opensource KING! Powerful, Fast, & Cheap! (Fully Tested)
The open-source community introduces GLM 4.5 and GLM 4.5 Air, two advanced large language models excelling in reasoning, coding, and agentic tasks, with competitive performance and cost-effective API pricing.
MAIN POINTS FROM TRANSCRIPT
- GLM 4.5 and GLM 4.5 Air are new models with 355 billion and 106 billion parameters, respectively.
- Both models support a 128k context length and are trained on 22 trillion tokens.
- GLM 4.5 ranks third in evaluation across 12 benchmarks, while GLM 4.5 Air ranks sixth.
- API pricing is cost-effective, with GLM 4.5 at $0.60 per million input tokens and $2.20 per million output tokens.
TAKEAWAYS
- Models feature hybrid thinking modes for deep reasoning and fast responses, similar to Alibaba's Quen 3 series.
- GLM 4.5 outperforms or rivals Claude GPT4.1 and Gemini 2.5 Pro in reasoning tasks.
- Open weights are available, allowing local use with smaller parameter sizes through Olama or LM Studio.
- Access models via API platforms like Open Router or chat with them using the Z AI chatbot.