OpenClaw has adopted DeepSeek’s new V4 Flash as the default model for the popular artificial intelligence agent, while the global tech community scrutinises the implications of the Chinese AI firm’s major update optimised to run on Huawei Technologies’ chips.

In an update on Sunday, OpenClaw said it had expanded its catalogue to include DeepSeek’s V4 Flash and its flagship V4 Pro models, and had integrated other capabilities such as Google Meet into the agent. OpenClaw also optimised how DeepSeek V4 models could remain consistent when performing multi-step tasks.

The move came two days after Hangzhou-based DeepSeek released its open-source V4 models, with the V4 Pro boasting 1.6 trillion parameters, making it the company’s biggest-ever model by that metric, while the smaller V4 Flash model has 284 billion parameters.

DeepSeek said on Friday that its V4 models were optimised for mainstream agent tools including Anthropic’s Claude Code, OpenClaw and Tencent Holdings’ CodeBuddy.

The DeepSeek V4 release drew global attention as its collaboration with Huawei, which announced “full support” from its Ascend chips and supernode systems for model inference, put China’s home-grown AI hardware to the test.