Frontier Models
Qwen3.8-Max Enters >2T Open-Weight Club With 2.4T Parameters
Alibaba's Qwen3.8-Max reaches general availability as the second model above 2 trillion parameters to pursue open weights after Moonshot's Kimi K3, with the full release scheduled for next week in the intensifying China frontier race.
Qwen3.8-Max is the most capable model in the Qwen family to date, scaling to 2.4 trillion parameters with 95 billion active parameters and set for open-weight release next week.
The general availability announcement for Qwen3.8-Max positions Alibaba at the forefront of open-weight frontier model releases. The model builds directly on the established Qwen series capabilities. Developers now await the weights release to begin integration work.
Chinese AI firms have accelerated large-scale model development in recent months. Open-weight strategies differentiate participants in this competitive space. Qwen3.8-Max enters as a direct response to earlier releases from peers.
What background led to the Qwen3.8-Max announcement?
Alibaba has iterated on the Qwen family through successive model generations focused on coding and reasoning tasks. The latest version extends this lineage with increased scale. Prior models in the series established benchmarks that the new release aims to surpass.
The broader industry shift toward open weights has influenced multiple organizations. Transparency in model architecture encourages external contributions and scrutiny. This environment supports faster iteration across the AI research community.
Moonshot AI set an early precedent with its Kimi K3 release. That model established the 3-trillion parameter open-source category. Subsequent entries now follow similar paths to maintain competitive parity.
What technical specifications define Qwen3.8-Max?
Qwen3.8-Max employs a mixture-of-experts design that activates only a subset of parameters during use. The total parameter count reaches 2.4 trillion while active parameters number 95 billion. This configuration balances performance with computational efficiency.
The architecture supports advanced performance in coding and collaborative tasks. Official descriptions highlight the model as the strongest in its family. Scale improvements enable more sophisticated output generation compared to earlier variants.
How does Qwen3.8-Max compare to Kimi K3?
Kimi K3 holds 2.8 trillion parameters and was the first open-source model to enter the 3-trillion class. Its full weights became available by July 27, 2026. Qwen3.8-Max follows with a comparable but slightly smaller total parameter count.
Both models target the high end of open-weight frontier capabilities. Release timing places Qwen3.8-Max as the immediate successor in the category. Differences in active parameter counts affect inference resource requirements.
| Model | Total Parameters | Active Parameters | Open Weight Status | Release Timeline |
|---|---|---|---|---|
| Qwen3.8-Max | 2.4 trillion | 95 billion | Scheduled next week | GA live now |
| Kimi K3 | 2.8 trillion | Not specified | Released | By July 27, 2026 |
What market and stakeholder implications follow from the release?
Access to open weights enables developers to fine-tune Qwen3.8-Max for specialized applications. Enterprises gain options for on-premise deployment without reliance on closed APIs. The move broadens participation in frontier model usage.
The China open-source frontier race gains another major participant. Additional firms may accelerate similar releases to remain relevant. Wider availability of these models supports ecosystem growth around customization and evaluation.
Researchers benefit from the ability to inspect and extend the model architecture. Community contributions can address specific limitations through targeted modifications. This dynamic fosters collaborative progress in AI capabilities.
What expert reactions have emerged around Qwen3.8-Max?
The Qwen research team issued statements detailing the model's capabilities and the significance of the open-weight decision. Emphasis was placed on the historic aspect of releasing a Qwen-Max class model. The announcement highlights ongoing evolution of the series.
Today, we are officially releasing Qwen 3.8-Max, the most capable model in the Qwen family to date. This also marks the first time we will open-source the weights of a Qwen-Max-class model — the open weights will be released next week.QwenTeam, Qwen research team
An official post from Alibaba Qwen described the model as one of the most powerful available. Compatibility with leading frontier models was noted, with positioning second only to Fable 5. These comments frame the release within the current competitive hierarchy.
What comes next for Qwen3.8-Max and related developments?
The open weights release is confirmed for the week following the general availability announcement. Users can prepare environments for download and initial testing. Subsequent community-driven evaluations will provide performance data.
Further iterations in the Qwen series may incorporate feedback from the open release. The competitive pressure from models such as Kimi K3 is expected to persist. Additional large-scale open-weight entries could appear from other organizations.
Stakeholders should monitor official channels for precise timing of the weights distribution. Integration planning can begin based on the published specifications. Long-term implications include expanded use cases across research and commercial domains.
- Review the official Qwen blog for the exact weights release date.
- Assess hardware requirements for running the 2.4T parameter model locally.
- Compare benchmark results against closed frontier models in target domains.
- Plan fine-tuning experiments once weights are accessible to the public.
Frequently asked
When will the open weights of Qwen3.8-Max become available?
The open weights of Qwen3.8-Max will be released next week following the general availability announcement.
How many parameters does Kimi K3 contain compared to Qwen3.8-Max?
Kimi K3 contains 2.8 trillion parameters while Qwen3.8-Max scales to 2.4 trillion total parameters with 95 billion active.
Sources
- Qwen — Qwen3.8-Max is the most capable model in the Qwen family to date with 2.4T parameters (95B active) and open weights releasing next week.
- Moonshot AI / Kimi — Kimi K3 is the world’s first open-source model in the 3-trillion-parameter class with 2.8 trillion parameters, full model weights released by July 27, 2026.
- Alibaba Qwen — Qwen3.8 is launching and going open-weight soon with a massive 2.4T parameters, compatible to leading frontier AI models, second only to Fable 5.