Qwen3.8-Max-Preview
PreviewAlibaba's largest model to date and the flagship of the new Qwen3.8 line — a 2.4-trillion-parameter, fully multimodal model that Alibaba positions just behind Anthropic's Fable 5 on overall performance (vendor internal evals; no independent third-party benchmarks yet). Launched Jul 19, 2026 as Qwen3.8-Max-Preview, available to developers via Alibaba's Token Plan subscription and the Qoder / QoderWork coding platforms. Architecture is presumed sparse-MoE; active-parameter count, context window, and pricing are undisclosed. Breaking from the API-only pattern of earlier Max models, the Qwen team says it will release open weights "soon," though no timeline, license, or full specs have been announced.
Specifications
- License
- Proprietary
- Weights
- Not released
- Architecture
- Mixture-of-Experts
- Parameters
- 2.4T
- Context window
- — tokens
- Max output
- —
- Knowledge cutoff
- —
- Price (in / out, $/M)
- —
- Modalities
- TextVisionCode
Benchmarks
No benchmark scores recorded yet. Spotted some? Submit a correction.
Vendor-reported figures are claims until independently verified. See methodology.