Alibaba's Qwen3.8-Max Challenges GPT-5.6 and Claude Fable 5 on Coding Benchmarks, Open Weights Coming August 10
Alibaba Debuts Qwen3.8-Max with Advanced Coding Capabilities
Alibaba's Qwen team officially released Qwen3.8-Max on August 3, 2026, a 2.4-trillion-parameter mixture-of-experts (MoE) model that challenges GPT-5.6 Sol and Claude Fable 5 on coding benchmarks, with open weights promised for the week of August 10, 2026. For broader context, explore our AI News.
Performance on Key Coding Benchmarks
Qwen3.8-Max demonstrated strong performance in coding evaluations. On Terminal-Bench 2.1, the model achieved a score of 86.6. This places it ahead of Claude Opus 4.8 and Claude Fable 5, both of which scored 84.6. Qwen3.8-Max was surpassed only by GPT-5.6 Sol's maximum reasoning mode, which recorded a score of 88.8.
In additional coding assessments, Qwen3.8-Max scored 1,668 on the Frontend Code Arena, trailing Claude Opus 5 by 37 points. The model also completed a 16-day coding project without human intervention and successfully optimized a chip design task involving over 500 steps.
Technical Specifications and Features
Qwen3.8-Max is built as a mixture-of-experts model, utilizing a vast 2.4-trillion-parameter architecture. Its ability to activate 95 billion parameters per query contributes to its processing efficiency. The model's 1-million-token context window allows for extensive input and output processing, supporting multimodal inputs including text, images, and video.
Key features of Qwen3.8-Max include:
- Function calling
- Structured outputs
- Code interpreter
- Web search capabilities
- Fine-tuning support
API Access and Pricing Structure
Qwen3.8-Max is available via API. The pricing structure is set at $2 per million input tokens and $6 per million output tokens. Additionally, cached input is priced at $0.25 per million tokens. This pricing model is positioned to be competitive within the market for large language models.
Upcoming Open Weights Release
Alibaba has announced plans to release open weights for Qwen3.8-Max and a smaller 27B variant, Qwen3.8-27B. This release is scheduled for the week of August 10, 2026. The availability of open weights will allow developers and researchers to access and utilize the model's architecture for various applications and further research.
Conclusion
Alibaba's Qwen3.8-Max represents a significant entry into the competitive landscape of large language models, particularly with its strong performance in coding benchmarks and multimodal capabilities. The model's release, coupled with its competitive API pricing and upcoming open weights, positions it as a notable option for developers and enterprises. The promised open weights release in mid-August 2026 will likely expand its adoption and integration into diverse AI applications.
Sources
Was this article helpful?
Found outdated info or have suggestions? Send us a note.
