Alibaba’s 2.4-Trillion-Parameter Qwen3.8-Max Claims Second Place Only to Anthropic
Alibaba has launched Qwen3.8-Max, a 2.4-trillion-parameter multimodal model it says is second only to Anthropic’s Fable 5 among frontier AI systems. The release, which sent Alibaba shares higher, intensifies the U.S.-China AI race and challenges both domestic rival Moonshot AI and Western leaders.
Alibaba has officially launched Qwen3.8-Max, a 2.4-trillion-parameter multimodal model it describes as its largest and most capable flagship to date — and one that it claims is “second only to Anthropic’s Fable 5” among the systems it has benchmarked. The announcement, made from Hangzhou on August 3, 2026, sent Alibaba’s shares rallying as investors bet on the Chinese tech giant’s ability to compete at the frontier of artificial intelligence. The model arrives amid an escalating U.S.-China AI arms race, just days after domestic rival Moonshot AI unveiled its own 2.8-trillion-parameter open-weight model, Kimi K3.
What happened
Qwen3.8-Max is built on the Qwen 3.5 foundation and uses a Sparse Mixture-of-Experts (MoE) architecture combined with a hybrid attention mechanism. It supports a context window of up to 1 million tokens, enabling long-horizon reasoning and large-document workflows. Alibaba says the model can handle text, images, video, and documents, making it its first multimodal model above 1 trillion parameters.
A preview version, Qwen3.8-Max-Preview, was first shown on July 19, 2026, at the World AI Conference in Shanghai. The preview has been accessible via Alibaba’s Token Plan subscription, as well as through its coding platforms Qoder and QoderWork, at just 10% of standard pricing. The final version is now available through APIs on Alibaba Cloud Model Studio for global developers.
Alibaba’s public messaging has been bold. On X (formerly Twitter), the Qwen team stated: “We believe it’s one of the most powerful models available today, compatible to leading frontier AI models, second only to Fable 5.” Bloomberg and other outlets have echoed that characterization, noting Alibaba’s claim that Qwen3.8-Max is “comparable to leading frontier AI models and second only to Anthropic PBC’s Fable 5.”
💡 Alibaba is positioning Qwen3.8-Max as a direct rival to Western frontier models, but its claims rest on internal benchmarks that have not yet been independently verified. The company has not published a full model card or benchmark table.
The model’s internal benchmarks reportedly show it outperforms the previous Qwen3.7-Max on coding, data analysis, and general office workflows. Externally, Alibaba states that Qwen3.8-Max ranks fifth in Text Arena and second in Vision Arena, two evaluation platforms for text and vision models. These rankings are used to substantiate its “frontier-comparable” claims.
Why it matters
The launch of Qwen3.8-Max comes at a critical moment in the U.S.-China AI competition. Chinese companies have been racing to close the gap with U.S. players such as OpenAI and Anthropic, and the release of Moonshot AI’s Kimi K3 — a 2.8-trillion-parameter open-weight model — just two days before the Qwen preview upended perceptions of China’s AI capabilities and roiled global technology stocks. Qwen3.8-Max is Alibaba’s answer, both to Kimi K3 and to Western frontier models.
Investors took notice. Alibaba’s shares rose following the unveiling, reflecting a market reassessment of Chinese tech competitiveness. The rally was also buoyed by Alibaba’s recent win as technology partner for Apple Intelligence in China, following Beijing’s approval of the partnership. That deal positions Alibaba as a key AI infrastructure provider in the world’s second-largest economy.
💡 The rapid succession of large-model launches from Chinese firms — Kimi K3 then Qwen3.8-Max — signals that the AI frontier is no longer a U.S.-only story. Global enterprises and developers now have credible alternatives from China, though independent validation remains pending.
What it means for business
For developers and enterprises, Qwen3.8-Max offers a practical alternative to Western models. Alibaba says the model is compatible with OpenAI and Anthropic protocols, allowing existing coding agents and client applications to switch to Qwen3.8-Max with minimal changes. The preview pricing at 10% of standard rates makes it a cost-effective option for experimentation.
The model is integrated into QwenWork, Alibaba’s all-in-one workplace AI agent platform, targeting office productivity, data analysis, coding, and research tasks. For businesses already using Alibaba Cloud, Qwen3.8-Max is a natural upgrade.
Perhaps most significantly, Alibaba has publicly committed to open-weight release “next week” — a shift from its previous “Max” flagships, which shipped as closed-weight models. The company says it intends to “open-weight it for everyone.” However, the license terms have not yet been specified, and independent analysts expect a restrictive, non-open-source license similar to other large corporate models.
💡 For developers, the promise of open weights is the real prize — it enables self-hosting, fine-tuning, and independent benchmarking. But the actual license will determine whether Qwen3.8-Max becomes a genuine open alternative or just another controlled API product.
What to watch next
The next week will be pivotal. Alibaba has said it will release model weights, and the community will be watching for the full benchmark suite, model card, and license details. Third-party evaluations will test the claim of being “second only to Fable 5.” Meanwhile, Qwen’s team has emphasized that the model is “still evolving daily” and that the latest version shows “broad gains and a big step up on web frontend.”
If the open-weight release delivers on performance, Qwen3.8-Max could reshape the competitive landscape for large language models — not just in China, but globally. For now, Alibaba has thrown down a gauntlet that Anthropic, OpenAI, and Moonshot AI will all have to answer.
Want automation like this for your business?
Get in touch and we'll show you exactly what's possible for your setup.