Alibaba launches a 2.4T-parameter Qwen3.8 Max preview, saying it rivals frontier AI models and is second only to Fable 5, planning to make it “open-weight soon”
Bloomberg
Context & Ripple Effects
Alibaba is extending a Qwen roadmap that moved from a more-than-1T-parameter Qwen3-Max preview to an open-weight Qwen3.5 release and a separate small-model series. The stated plan to open-weight Qwen3.8 Max brings its largest model closer to that distribution strategy.
The announcement matters because Alibaba is pairing a frontier-performance claim with prospective weight availability, rather than positioning scale only as a closed cloud service.
First-order effects
Developers and enterprise AI teams may gain another prospective open-weight option at the frontier end of Alibaba's Qwen lineup, subject to the promised release and independent performance validation.
Alibaba raises the profile of its AI platform by using Qwen3.8 Max's claimed standing against frontier models to market its model and cloud ecosystem.
Second-order effects
Other open-weight model providers face greater pressure to compete on performance, deployment flexibility, and the cost of running large models—not only on parameter scale.
A released model of this size would shift more value toward hosting, fine-tuning, evaluation, and inference infrastructure, where users need support beyond raw weights.
Third-order effects
If frontier-capable models increasingly become open-weight, model buyers gain leverage to mix self-hosted and managed deployments rather than relying solely on proprietary APIs.
The pattern would sharpen the split between companies that can fund frontier training and the broader ecosystem that commercializes models through infrastructure and services.
The trend: Frontier AI competition is increasingly being fought through both model capability claims and the distribution choice between closed access and open weights.
...We believe it's one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba's Token Plan, Qoder, and QoderWork. [image]
Qwen3.8-Max Preview is now available for early access! This is our first trillion-parameter multimodal model. Based on my own experience, it not only delivers multimodal understanding that is competitive with—and in many cases ahead of—today's leading proprietary models, but
Open weights models really accelerated! > Qwen-3.8 (2.4T), “second only to Fable 5” > Kimi-K3 (2.8T), “only behind Claude Fable 5 Max and GPT-5.6 Sol Max” > GLM-5.2 (753B): remarkably efficient and capable. The demand we're seeing from Databricks customers is wild. This is sup…
One of the horrible effects of Doomerism has been to reduce vital scientific communication, thus making US labs less competitive with time and giving an advantage to China (which is, paradoxically, far freer in this respect).
Regardless of what you think the answer should be, the inherent tension between a growing US regulatory/approval regime for frontier closed models & the lack of one for open models is going to need to be resolved in some way or another in the near future, with large consequences.
Qwen's biggest models have never been open weight recently. Something is changing. You can imagine it like... Xi: no you're not allowed to keep your best model closed anymore, we weren't succeeding enough. A new era of competition for intelligence (if the benchmarks hold up).
Holy, Qwen 3.8 supposedly ahead of GPT-5.6 and only slightly behind Fable 5! - 2.4t Parameters - Open Source / Open Weight - full release soon, already available for testing as Qwen 3.8 max-Max-Preview What the frick, such insane release on a sunday?! The gap between US closed …
Qwen 3.8 just got announced!. an open weight model, 2.4 trillion params, that alibaba's qwen team is calling second only to fable 5. if that holds, the number two model on earth, and the weights are coming to you, not a rented endpoint. every time this lab moves, it adds ten [ima…
1. Open source models are le bad and also decel 2. Government should make them banned 3. You should rent tokens from our price fixed oligopoly forever and be under threat of us removing access for any reason 4. My text is all lowercase because it's the vogue thing 5. No non o…
how dare they give us the model for free after stealing from america! the audacity! the horror! such foolish decels! such blindness to agi! such short-sightedness to not withhold these dangerous capabilities from the evil plebs and nefarious state actors who will now run 2.4T [im…
why can the china labs build glm-5.2, kimi k3, and many more to come? it is because of the openness. not just the open weights but the whole ecosystem. most of the work done in the china labs is carried by interns. i met brilliant undergrad and graduate interns who deeply und…