Alibaba’s Qwen3.8-Max rivals Kimi K3 and DeepSeek V4 Flash
Qwen3.8-Max (2.4T) matches top models at $2/M input tokens—with open weights next week
Alibaba’s Qwen team has launched **Qwen3.8-Max**, a 2.4-trillion-parameter open-weight model that rivals proprietary giants like Kimi K3 and DeepSeek V4 Flash across standard benchmarks. Unlike closed alternatives, this model prioritizes transparency and community access, with weights scheduled for release next week. The model’s strength lies in software and coding tasks, where it reportedly outperforms competitors in head-to-head evaluations.
Pricing is aggressively competitive: $2 per million input tokens, $6 per million output tokens, and an optional implicit caching fee of $0.25 per million tokens. A smaller variant, Qwen3.8-27B, is also slated for open-weight release. This move underscores Alibaba’s commitment to democratizing high-performance AI, offering enterprises and developers cost-effective alternatives without sacrificing quality.
- Qwen3.8-Max (2.4T params) matches Kimi K3 and DeepSeek V4 Flash on benchmarks, excelling in coding
- Open weights drop next week; pricing starts at $2/M input tokens with optional caching
- Alibaba also releasing Qwen3.8-27B (open-weight) soon
Why It Matters
Open-weight models like Qwen3.8-Max give businesses cost-effective, high-performance alternatives to proprietary AI.