Chinese open-source AI models join AI agent OpenClaw

Chinese open-source AI models, including Moonshot AI’s Kimi and MiniMax, are being integrated into OpenClaw, a popular global AI agent, due to their perceived cost efficiency and performance, industry experts say.

OpenClaw announced on January 30 that it offers Moonshot AI’s Kimi K2.5 and Kimi Coding for free, while supporting MiniMax, another Chinese foundation model developer.

Analysts note that Chinese open-source models are favored for their affordability, with Moonshot AI’s Kimi costing around US$0.6 per million input tokens, significantly lower than US models like Anthropic’s Claude Opus 4.5, which charges US$5 per million input tokens.

Industry insiders suggest that many users run these models autonomously, often incurring higher token consumption and unexpected costs.

OpenClaw, which launched in late 2025, is known for its ability to connect to various online services and perform diverse tasks, gaining attention in China and abroad.

🔗 Source: South China Morning Post

🧠 Food for thought

Implications, context, and why it matters.

OpenClaw’s design makes its users sensitive to AI model costs

OpenClaw is free and open-source. Many deployments still need a third-party AI model, so API calls become the regular expense 1.
The tool can run on a user’s own machine such as Mac, Windows via WSL2, or Linux. It can also connect to online services based on setup, and it reads as a technical self-hosted project 1.
Always-on automation such as persistent memory and a “heartbeat” feature can keep tokens flowing. Paid APIs can rack up charges faster than expected 1.
One developer reported a $623 API bill in a single month. That example helps explain why cheaper models matter in this community 1.

OpenClaw’s model additions track sharper competition on price

Industry experts say OpenClaw is adding Chinese open-source models such as Moonshot AI’s Kimi and MiniMax for lower costs plus solid performance 2.
OpenClaw said it provides Moonshot AI’s Kimi K2.5 and “Kimi Coding” for free, and it supports MiniMax 2.
Analysts say these models often win on “value for money.” They also warn that autonomous agents can burn through tokens and inflate service bills 2.
In a wider AI pricing war, premium models can cost far more than alternatives. High-volume agent workloads can squeeze providers with higher prices 3.