Executive Summary
- •Xiaomi released MiMo-V2.6-Pro, claiming the top spot globally for open-weight AI models.
- •The MIT-licensed model scored 46 on the Intelligence Index and costs just $0.435 per million input tokens.
- •The system was trained entirely on Chinese silicon, demonstrating massive reinforcement learning scale without Nvidia dependencies.
Community Sentiment
Key Developments & Data
Xiaomi's MiMo-V2.6-Pro scored a 46 on Artificial Analysis’ Intelligence Index, edging out xAI’s Grok 4.6 and Google’s Gemini 3.8 Flash to become the top-performing open-weight model globally.
The open-weight model is MIT-licensed and priced aggressively at $0.435 per million uncached input tokens and $0.87 per million output tokens.
Engineers completed the reinforcement learning phase across 30 large steps covering roughly 750,000 trajectories in under six days, costing approximately $2.62M for the Pro tier.
To prevent reward hacking in coding tasks, the training environment stripped build artifacts, removed Git history, and deployed a specialized "hack agent" to close loopholes before the final run.
Industry observers noted the capability gains were achieved using Chinese silicon rather than Nvidia GPUs.
"In an era when compute is brutally scarce, we still chose to dedicate a team of several dozen people to one goal over an extended period: scaling up RL." — Fuli Luo
Zubiqo Intelligence Briefing
Get the unfiltered signal before markets open.
Top tech breakthroughs, venture funding, and market moves—synthesized into a 2-minute morning read. Zero PR fluff.
✓ 100% Free•✓ 1-click unsubscribe•✓ No spam ever
Zubiqo Strategic Assessment
Primary Impact
Enterprise software development and proprietary AI API markets, as aggressive open-weight pricing undercuts established Western frontier models.
Strategic Shift
The separation of advanced reinforcement learning from Nvidia hardware dependencies, proving massive agentic scaling is viable on domestic Chinese silicon.
The Ripple Effect
Western AI labs will be forced to slash API pricing for agentic workloads to compete with Xiaomi's baseline rate of $0.435 per million input tokens.
This intelligence assessment is generated by Zubiqo's AI for informational purposes only.
