Alibaba's new Qwen3.8 Max model features 2.4 trillion parameters and is available for testing, with claims of rivaling Fable 5 on internal benchmarks. This follows Kimi K3, a 2.8 trillion parameter model, hitting GPU capacity within 48 hours of its release.
The release of massive models like Qwen3.8 Max and Kimi K3 highlights the growing compute demand for inference, even as their production viability remains unproven.