Alibaba Launches 2.4T Parameter Qwen3.8-Max with 1M Context and Multimodal Support
togethercompute · x · 2026-08-13
Together AI announced the deployment of Tongyi Qianwen's latest flagship model, Qwen3.8-Max (Qwen3.8-2.4T-A95B). Featuring a sparse Mixture-of-Experts architecture with 2.4 trillion total parameters, it is the largest model released by the Qwen team to date and their first multimodal model exceeding one trillion parameters.
Key Features:
- Massive Context: Natively supports a 1 million token input window with up to 128K output tokens, ideal for processing entire codebases or extensive documents.
- Multimodal & Reasoning: Processes text and image inputs. Always-on thinking mode offers three adjustable reasoning effort levels (low, high, xhigh).
- Agentic Workflows: Optimized for coding and long-horizon agentic tasks, supporting task state dispatch, self-testing (unit/E2E), and structured outputs across loops.
The model was officially announced at the World AI Conference (WAIC) in Shanghai on July 19, 2026, with Alibaba committing to an open-weight release.
More from coding & agent
- DIRECT: retrieval-based next-question suggestions for conversational agents — burkov · 2026-10-16
- Claude Code and Codex now autocomplete the user's next action — tokenbender · 2026-10-16
- Hermes Agent ships official Reverse Engineer Anything plugin with one-command setup — Teknium · 2026-10-16
- MLExpert: 35B MoE Agent Trained to Autonomously Handle Machine Learning Engineering — Shangheng Du · 2026-10-16
- Cowork-Gym: Alibaba Synthesizes 64.9K Agent Work Environments, Boosting Qwen by 8.7 Points — Anni Zou · 2026-10-16
- MMSearch-OPD: Joint RL and On-Policy Distillation for Long-Horizon Multimodal Search Agents — PekingUniversity · 2026-10-16