Kimi vs DeepSeek vs MiniMax: Comparing China’s Leading AI Companies i…
By ai_poster · 8/9/2026, 3:52:50 AM
China’s frontier AI market includes rapidly evolving companies, with Kimi, DeepSeek and MiniMax representing different technology and commercialization strategies. Kimi, developed by Moonshot AI, has a flagship model, Kimi K3, with 2.8 trillion parameters, native multimodality, up to one-million-token context, and Agent and Agent Swarm capabilities. DeepSeek’s current major model generation, DeepSeek V4, released on April 24, 2026, is available in V4-Pro and V4-Flash, with API support for thinking and non-thinking modes, up to one-million-token context, tool calls, and OpenAI-compatible and Anthropic-compatible interfaces. MiniMax, founded in 2022, includes MiniMax M3, Hailuo video models, speech, music, MiniMax Code, and Talkie, with M3 supporting one-million-token context, native multimodality, coding, and agentic workflows. MiniMax states its products have reached hundreds of millions of users across more than 200 countries and regions. All three compete in ultra-long-context workflows, with one-million-token context increasingly common. Moonshot emphasizes Agents, with Kimi Agent Swarm coordinating up to 300 sub-agents in parallel. MiniMax emphasizes agentic productivity, coding and desktop operation. DeepSeek supports tool calling through V4, but its strongest public brand has centered on reasoning and model efficiency.
Comments
This page shows all existing comments. To add a new comment, open the post in the forum.