Introducing Open-weight models
Run Kimi, Qwen, DeepSeek, and GLM in chat, at a fraction of US lab prices, served on US infrastructure.
The problem
If you run SEO yourself, or you are the lean growth person at a B2B SaaS, the expensive US lab models chew through credits. You start saving Claude Opus, Claude Fable, and GPT-5.6 Sol for "the real work" and you run fewer briefs, fewer competitor checks, and fewer draft loops than the backlog needs.
That is a budget problem, not a skill problem. You already know what to do. You cannot afford flagship prices on every turn of the agent.
Open-weight models from Moonshot, Alibaba, DeepSeek, and Z.AI have closed most of that quality gap on public leaderboards. The leftover objection has been hosting. Nobody wants to paste a brand brief into an API they cannot explain to legal.
What's new
Those models are now in Zerply chat, in the same picker as Claude, GPT, and Gemini. Open Zerply Agent, click the model name next to the send button, and search for Kimi, Qwen, DeepSeek, GLM, or Grok.
Open-weight means the model is not locked inside one US lab. US providers can host it, which is how Kimi, Qwen, DeepSeek, and GLM show up in Zerply without sending your prompts to a China-hosted API. You still pay in the same AI Agent credits.
What is in the picker:
- Kimi K3 (Moonshot)
- Qwen 3.8 Max, Qwen 3.7 Plus, and Qwen 3.7 Flash (Alibaba)
- DeepSeek V4 Pro
- GLM 5.3 and GLM 5.2 (Z.AI). 5.3 is the latest, same price as 5.2, stronger on long agent work
- Grok 4.6 (xAI)
Claude Opus 5 is already in that list. Keep it for the jobs that earn flagship spend.
On Artificial Analysis's Intelligence Index, Kimi K3 scores 57 with max reasoning. Claude Fable 5 is 60. GPT-5.6 Sol is 59. That is a few points, not a generation. DeepSeek V4 Pro, GLM, and the Qwen 3.7 line cost a fraction of Opus and Sol, so volume work does not have to live on the flagship.
What you can do now
- Spend credits on more work: Run cluster briefs, competitor checks, and first drafts on DeepSeek, GLM, Qwen, or Grok, and keep Opus or Fable for the jobs that earn the extra spend.
- Stay on US infrastructure: Use the Chinese labs' models without sending the brief to a China-hosted API.
- Switch without leaving chat: The same picker is on new blog and landing page drafts, so the model follows the job.
Try it now
Start free trial, open Zerply Agent, and click the model name next to the send button. Search for Kimi, Qwen, DeepSeek, GLM 5.3, or Grok and run the next brief on it.
Written by
Founder, Zerply.ai & Wittypen
Preetesh Jain is the Founder of Zerply.ai and Wittypen. He specializes in SEO, Answer Engine Optimization (AEO), AI search visibility, content marketing, and product development. Through his work building AI-powered marketing platforms, he helps businesses improve their organic presence across Google, ChatGPT, Perplexity, Claude, and other emerging discovery channels. He regularly writes about AI search, organic growth, content strategy, and the future of digital marketing.
