Qwen Launches QwenCloud Platform, Reveals 2.4T-Parameter Qwen3.8-Max Flagship

Alibaba_Qwen · x · 2026-09-04

Qwen officially launched QwenCloud (qwencloud.com), an AI-native models/tools/apps platform. Highlights: Qwen3.8-Max, its most capable flagship yet—a native vision-language MoE model with 2.4 trillion parameters, 1M context, 131.1K max output, priced at $2/M input and $6/M output; Wan-Video for controllable video generation at $0.068–0.28/second; and Qwen3.8-Flash with 1M-token context and OpenAI/Anthropic API compatibility for tools like Claude Code and Codex.

Original post →

More from Models

Models channel →