AI
Open models recap: more on Kimi K3, Qwen 3.8, Xi's WAIC speech, distillation, the open
In a recap of the open-source AI ecosystem, experts discuss how Chinese models like GLM 5.2 and Kimi K3 are demonstrating capabilities comparable
Key takeaways
- GLM 5.2 is at a Claude Sonnet-ish level in terms of capabilities.
- The weights for GLM 5.2 have not been released yet.
- Buying external data is helping open models catch up to closed ones.
In a recap of the open-source AI ecosystem, experts discuss how Chinese models like GLM 5.2 and Kimi K3 are demonstrating capabilities comparable to Anthropic's Claude Sonnet. While GLM 5.2 is highly fast and integrated into agentic coding workflows, its weights have not yet been publicly released, highlighting a different rollout regime compared to other open-source models.
In their words
“I would say you could probably go really far for a lot of the work with Kimi K3 as the main agent and GLM for sub agent work.”
“These are not just benchmaxxed distilled IP theft models. These are like genuinely good models that people are comparing to on their internal trading benchmarks.”
By the numbers
- 200
- Tokens per second delivered by the GLM API
- 300
- Tokens per second delivered by the GLM API
How it unfolded
- Kimi K3 was released
Turn stories like this into views
Ravenclip finds the AI news, makes the video, and posts it before attention moves on.
Common questions
- What happened with Kimi K3?
- GLM 5.2 is at a Claude Sonnet-ish level in terms of capabilities.
- Where can I read the original report?
- Read the full report at interconnects_ai.
