Qwen 3.8 Released : Better than Kimi K3?

2026-07-19 · Source: Machine Learning on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

Alibaba has officially announced Qwen 3.8, a new flagship large language model featuring a massive 2.4 trillion parameters, with an open-weight release anticipated soon. The company claims Qwen 3.8 is competitive with today's leading frontier AI models, ranking just behind Claude Fable 5 in overall capability. Developers can already experiment with Qwen3.8-Max-Preview through Alibaba's Token Plan, Qoder, and QoderWork, with subscription options ranging from a \$6 per month Lite Plan to a \$68 per month Pro Plan. While Moonshot AI's Kimi K3 reportedly has 2.8 trillion parameters, Qwen 3.8's upcoming open-weight status offers significant flexibility for deployment, fine-tuning, and customization, potentially making it a preferred choice for enterprises and researchers despite its slightly smaller scale.

Key takeaway

For Machine Learning Engineers evaluating frontier LLMs for enterprise deployment, Qwen 3.8's upcoming open-weight release presents a compelling alternative to larger, closed models like Kimi K3. Its 2.4 trillion parameters, combined with the flexibility to fine-tune and deploy on your own infrastructure, could significantly reduce vendor lock-in and enhance customization. You should explore the Qwen3.8-Max-Preview now to assess its real-world performance for your specific applications.

Key insights

Alibaba's Qwen 3.8, a 2.4 trillion parameter model, is set for an open-weight release, emphasizing deployment flexibility.

Principles

• AI model performance is not solely determined by parameter count.
• Open-weight models enable greater customization and deployment freedom.
• Continuous evolution models deliver more frequent capability enhancements.

In practice

• Test Qwen3.8-Max-Preview via Alibaba's Token Plan.
• Deploy open-weight models on private infrastructure.
• Customize models for specific enterprise applications.

Topics

• Qwen 3.8
• Large Language Models
• Open-weight Models
• Alibaba
• Kimi K3
• Model Benchmarking

Best for: CTO, VP of Engineering/Data, AI Engineer, AI Scientist, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• Alibaba's Qwen takes on Kimi K3 with open-weight Qwen 3.8, says model is "second only to Fable 5" — Alibaba's Qwen 3.8, a 2.4T parameter multimodal model, challenges Kimi K3 with open weights and claims near-Fable 5 performance.
• Qwen3.8: What Hardware Will You Need to Run Alibaba’s 2.4T Model? — Alibaba's 2.4-trillion-parameter Qwen3.8 model, planned for open release, offers unprecedented scale but is too large for typical local deployment.
• 😸 Alibaba’s 2.4T Qwen joins the AI race — Alibaba's 2.4T Qwen3.8-Max model, promised as open-weight, intensifies the competition between proprietary and accessible frontier AI.
• BREAKING: China's Kimi K3 Just Dropped a 2.8 Trillion Parameter Open AI Model — Kimi K3 introduces a 2.8T-parameter open MoE model with a 1M-token context, setting a new scale benchmark.
• Alibaba unveils 2.4T-parameter Qwen3.8 AI model — Alibaba's 2.4T-parameter Qwen3.8 model is launching with a new open-weight strategy, despite unverified performance claims.
• Qwen 3.8 Next First Test | Coding, Game Dev, Frontend | 2.4T Parameter Open Weight model by Alibaba — Alibaba's 2.4T parameter Qwen 3.8 Max Preview model shows inconsistent coding performance, often struggling with complex tasks and exhibiting slow "thinking."

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning on Medium.