MiniMax: M3 Pro Model Parameters Expected to Reach ~3T, Large-Scale Domestic Computing Cluster Coming Soon
Bubble signal: 0 · NeutralRelevance 45ModelsChina
Source: IT Home · Published 2026-08-26
Summary
MiniMax revealed in its interim earnings call that the M3 Pro model is expected to have around 3T parameters and will expand reinforcement learning training. The company is adapting domestic chips and will soon launch a large-scale domestic computing cluster. It also reported H1 FY2026 revenue up 283.1% YoY, but net loss still at $358 million.
Bubble analysis
This news mainly concerns model development and domestic computing cluster construction, which is AI industry progress but weakly tied to the investment bubble, as it lacks specific capex or valuation figures.
#minimax#model#domestic-compute
Read original ↗Related signals
linked by shared tags
- Trained on All-Domestic Computing Power with 293B Parameters! iFlytek Spark X2.5 Arrives, with Hands-On Test#model
- AI Video Generation Crosses Real-Time Threshold, Enabling Continuous Streams and Interactive Stories#minimax
- OpenAI Applied Research Head Discusses Spatial Reasoning, Says Astra Model Has Closed One of Humanity's Few Advantages#model
- LLM Vendors Rush to Sell Tokens, Kimi and MiniMax Reportedly to Open Stores on Tmall#minimax
- China's Open-Source Models Are Becoming the Global AI Benchmark#minimax