Alibaba Qwen3.8 claims second place behind Anthropic Claude Fable 5
Synopsis
Alibaba has previewed Qwen3.8-Max with 2.4 trillion parameters, claiming it trails only Anthropic's Claude Fable 5 — a bold ranking announced at WAIC 2026 in Shanghai, days after Moonshot AI's Kimi K3 set a record as the world's largest open-source model at 2.8 trillion parameters.
Key Takeaways
Alibaba Group Holding announced Qwen3.8-Max-Preview on 19 July 2026 at the World AI Conference (WAIC) in Shanghai .
The model carries 2.4 trillion parameters , which Alibaba says makes it 'second only' to Anthropic 's Claude Fable 5 .
Qwen3.8-Max-Preview is available on Alibaba 's Token Plan subscription and its Qoder and QoderWork agentic platforms.
Moonshot AI unveiled Kimi K3 on 17 July 2026 with 2.8 trillion parameters , currently the world's largest open-source AI model.
Multiple Chinese AI labs — including Zhipu ( GLM-5.2 ) and DeepSeek — are simultaneously pushing frontier model scale at WAIC 2026 .
Alibaba Group Holding has unveiled Qwen3.8-Max-Preview, its next-generation AI model packing 2.4 trillion parameters, declaring it 'second only' to Anthropic's Claude Fable 5 among frontier models currently available. The preview was announced via an official post on X on Sunday, 19 July 2026, coinciding with China's premier AI gathering, the World AI Conference (WAIC) in Shanghai.
What Alibaba is claiming
According to the company, Qwen3.8-Max-Preview is 'one of the most powerful models available today, comparable to leading frontier AI models, second only to Fable 5' — a reference to Anthropic's most capable widely released model to date. The preview version has been made available on Alibaba's Token Plan subscription service, as well as its agentic coding platforms Qoder and QoderWork.
Parameters are a standard machine-learning measure of AI system complexity; higher figures generally indicate stronger reasoning and generation capabilities. At 2.4 trillion parameters, Qwen3.8 places Alibaba firmly in what the company describes as the 'multi-trillion parameter' era of AI development.
The competitive backdrop
The announcement arrives just days after Beijing-based Moonshot AI unveiled Kimi K3 on Thursday, 17 July 2026, an open-source model with 2.8 trillion parameters — currently the largest open-source AI model in the world. Together, the two releases signal a rapid acceleration in Chinese AI model scale at a moment when the global frontier is being contested at the multi-trillion parameter level.
Other Chinese AI players including Zhipu, which recently released GLM-5.2, and DeepSeek are also pushing parameter counts and benchmark scores, intensifying the domestic race alongside the broader US-China tech competition.
Why it matters
WAIC 2026 in Shanghai has become the stage for a string of high-stakes AI model announcements, underscoring how Chinese developers are converging on — and in some cases matching — Western frontier models. Alibaba's positioning of Qwen3.8 as second only to Claude Fable 5 is a direct challenge to the perceived hierarchy of global AI capability.
For enterprise customers evaluating AI infrastructure, the availability of Qwen3.8-Max-Preview on subscription and agentic platforms means competitive-grade models are now accessible at scale through Alibaba's cloud ecosystem.
What's next
The preview designation suggests a full production release of Qwen3.8 is forthcoming, with performance benchmarks expected to face independent scrutiny from the research community. As Chinese and US AI labs continue to trade headline parameter counts, the race to the next capability threshold — and the compute infrastructure required to reach it — will remain a central battleground in global AI development.
Point of View
Not an independently verified benchmark result — a distinction that mainstream coverage is glossing over. What the announcement does confirm, however, is that Chinese AI labs have crossed the multi-trillion parameter threshold en masse, compressing what was once a multi-year capability gap with Western frontier labs into months. The clustering of trillion-parameter releases at WAIC 2026 — Moonshot's Kimi K3 at 2.8 trillion, Alibaba's Qwen3.8 at 2.4 trillion — suggests Chinese developers are now competing on raw scale as much as on efficiency, a strategic pivot from the DeepSeek-era emphasis on doing more with less compute. The deeper story is about who controls the inference infrastructure: Alibaba's immediate deployment of Qwen3.8 on Token Plan and agentic platforms signals that the real prize is locking in enterprise customers before the next capability leap arrives.
NationPress
20 Jul 2026
Frequently Asked Questions
What is Alibaba Qwen3.8 and when was it announced?
Qwen3.8-Max-Preview is Alibaba Group Holding's next-generation large language model, announced on 19 July 2026 at the World AI Conference (WAIC) in Shanghai. It features 2.4 trillion parameters and is the preview version of the most powerful model in Alibaba's Qwen family.
How does Qwen3.8 compare to Anthropic's Claude Fable 5?
According to Alibaba, Qwen3.8-Max-Preview is 'second only' to Anthropic's Claude Fable 5 among frontier AI models currently available. This ranking is Alibaba's own claim and has not yet been independently verified by third-party benchmarks.
Where can users access Qwen3.8-Max-Preview?
Alibaba has made Qwen3.8-Max-Preview available on its Token Plan subscription service and on its Qoder and QoderWork agentic coding platforms, as stated in the company's official post on X.
What is Moonshot AI's Kimi K3 and how does it relate to Qwen3.8?
Kimi K3 is an open-source AI model unveiled by Beijing-based Moonshot AI on 17 July 2026, carrying 2.8 trillion parameters — currently the largest open-source AI model in the world. Both Kimi K3 and Qwen3.8 were announced within days of each other at WAIC 2026, reflecting a rapid escalation in Chinese AI model scale.
Why are Chinese AI companies releasing multi-trillion parameter models now?
Chinese AI developers appear to be entering a new phase of frontier model development focused on raw scale, following an earlier period — exemplified by DeepSeek — that prioritised efficiency. The WAIC 2026 conference in Shanghai has served as a launchpad for multiple high-parameter releases, signalling intensified competition with Western labs in the global AI race.
