Alibaba unveils its largest AI model yet, DeepSeek's latest model is ultra-low cost
Alibaba unveiled Qwen3.8-Max, its largest AI model, with 2.4 trillion parameters and a mixture-of-experts design using 95 billion parameters at a time. The model is due next week and helped Alibaba shares rise about 7% in Hong Kong. Research firm Artificial Analysis said DeepSeek’s V4-Flash is far cheaper, at about 3 cents per test versus higher costs for peers.
How this was made
The 30-second read
Why it matters
For traders, the actionable element is Alibaba’s same-day equity reaction to a specific model unveiling, plus a stated next-week release window. For DeepSeek, the article provides a cost benchmark that may affect competitive expectations for AI spending, but DeepSeek is not the subject of the brief’s ticker extraction because the system rules require the article’s subject company for ticker inclusion.
Market read
Alibaba’s AI model launch is tied to a reported 7% Hong Kong share jump and a next-week release schedule, making it a near-term sentiment and momentum catalyst.
What to watch
The article does not quantify Alibaba’s inference cost, customer adoption, or monetization timeline, so the next-week release could disappoint if it lacks commercial traction.
Background
The piece frames two AI developments: Alibaba’s Qwen3.8-Max launch and DeepSeek’s V4-Flash cost advantage, both illustrating rapid frontier progress and aggressive pricing.
Ticker impact
Alibaba unveiled Qwen3.8-Max, a 2.4T-parameter model due next week, and the stock jumped 7% in Hong Kong trade.
Likely supports continued momentum in the near term, but follow-through depends on broader AI monetization signals beyond leaderboard performance.
The article provides a same-day share reaction (up 7% in Hong Kong) plus concrete model specs (2.4T parameters, mixture-of-experts, 1M tokens) and a release timeline next week, which can drive incremental trading interest.
Market effects
Reinforces competitive pressure in China’s frontier-model race, potentially intensifying cost and performance benchmarks for AI infrastructure and cloud demand.
Supports risk-on sentiment for Chinese tech equities tied to AI leadership narratives.
Highlights global benchmark comparisons (including Anthropic/OpenAI) that can influence investor expectations for AI unit economics worldwide.
Counterpoint
Leaderboard ranking and parameter counts may not translate into revenue; the market may be over-discounting near-term AI capability announcements.
Key entities
- companyAlibaba
Unveiled Qwen3.8-Max, described as its largest and most capable AI model to date, with 2.4T parameters and mixture-of-experts design.
- productQwen3.8-Max
Alibaba’s new AI model, due for release next week, claimed to handle text, images, and video and process up to 1M tokens.
- companyDeepSeek
Released V4-Flash, described as ultra-low cost versus other well-known models based on Artificial Analysis benchmarks.
- productV4-Flash
DeepSeek model with benchmark pricing cited as $0.14 per million input tokens and $0.28 per million output tokens.
