Hopp til hovedinnhold
Fredag 24. april 2026AI-nyheter, ferdig filtrert for ledere
SISTE:
DeepSeek åpner V4 Preview med 1M kontekst og API-kompatibilitetOpenAI lanserer GPT-5.5 for ChatGPT og CodexAnthropic og Amazon utvider AI-alliansen med 5 GW kapasitet og ny investeringDeepSeek åpner V4 Preview med 1M kontekst og API-kompatibilitetOpenAI lanserer GPT-5.5 for ChatGPT og CodexAnthropic og Amazon utvider AI-alliansen med 5 GW kapasitet og ny investering
Xiaomi Silently Drops a Free 1-Trillion-Parameter Model on OpenRouter
Breaking
CIOChinaOpenRouterLLMXiaomi

Xiaomi Silently Drops a Free 1-Trillion-Parameter Model on OpenRouter

JH
Joachim Høgby
22. mars 202622. mars 20264 min lesingKilde:

No press release. No fanfare. Just a listing on OpenRouter named "Hunter Alpha" with specifications that stopped the entire AI community cold: 1 trillion parameters, 1 million token context window, agent-focused architecture. Free to use.

It happened on March 18, 2026. And nobody knew who was behind it.

The AI community immediately began speculating. The specs looked suspiciously like what Chinese media had been reporting about DeepSeek V4. Developers tested the model and found it identified itself as "a Chinese AI model primarily trained in Chinese" with a knowledge cutoff of May 2025, exactly the same as DeepSeek's own models.

But it wasn't DeepSeek.

The model was MiMo-V2-Pro, developed by Xiaomi's AI division MiMo under the leadership of Luo Fuli, a former DeepSeek researcher. A trillion-parameter model from a smartphone manufacturer, free for anyone to use, built by someone who jumped from one of China's leading AI laboratories.

This is one of the most surprising AI moments of 2026 so far.

What does this mean?

Two years ago, trillion-parameter models were the exclusive domain of the world's largest technology companies with billions of dollars in infrastructure. Now a Chinese technology company best known for affordable smartphones is launching a model at that scale, at no cost, available via an open platform.

Several things make this launch remarkable:

The scale is real. MiMo-V2-Pro competes directly with leading Western frontier models on key benchmarks. This isn't a marketing exercise, it's a technical achievement.

The distribution strategy is aggressive. By going directly to OpenRouter and making the model free, Xiaomi bypasses normal channels and announcement cycles. They said nothing — they just put it out there.

The Chinese AI industry is maturing rapidly. Following the DeepSeek wave in early 2025, we're now seeing Chinese laboratories produce frontier models at a steady pace. Xiaomi won't be the last company to surprise the market this way.

For CIOs and technology leaders, the message is clear: the geopolitical situation makes AI development more complex, but the technological competition is now genuinely global. Chinese models can no longer be ignored.

Tencent launched a tool the same week to integrate WeChat with its AI agent platform, further underscoring that China's technology companies are now fighting hard for control of AI agent infrastructure.

📬 Likte du denne?

AI-nyheter for ledere. Kuratert av en CIO som bygger det selv. Daglig i innboksen.

Relaterte saker

Meta velger AWS Graviton for agentisk AI i stor skala
CIOInfrastructure

Meta velger AWS Graviton for agentisk AI i stor skala

Akkurat nå4 min lesing
Åpne saken
Meta taps AWS Graviton to scale agentic AI
CIOInfrastructure

Meta taps AWS Graviton to scale agentic AI

Akkurat nå4 min lesing
Åpne saken
DeepSeek åpner V4 Preview med 1M kontekst og API-kompatibilitet
Breaking
CIOOpen Source

DeepSeek åpner V4 Preview med 1M kontekst og API-kompatibilitet

Akkurat nå4 min lesing
Åpne saken