Alibaba released a 2.4 trillion parameter AI model Monday, lifting its Hong Kong shares 7%, as DeepSeek undercut Anthropic's Claude Fable 5 by more than 100 times on cost.
Key Points:
- Alibaba's Qwen3.8-Max carries 2.4 trillion total parameters and activates about 95 billion per query, with open weights due next week.
- Hong Kong shares closed 7% higher, while New York listed receipts rose about 4% before the open.
- DeepSeek's V4-Flash costs roughly 3 cents per benchmark test, against $3.15 for Claude Fable 5.
Alibaba Qwen3.8-Max Launch Lifts Shares
The Hangzhou company released Qwen3.8-Max on Monday and called it the most capable system in its Qwen family to date. The model carries 2.4 trillion total parameters, the settings a system learns from data, and activates roughly 95 billion for any single query. Its Hong Kong shares closed 7% higher. American depositary receipts rose about 4% in premarket trading in New York.
Qwen3.8-Max handles text, images and video, and it takes in up to 1 million tokens at a time. Alibaba built the model on its earlier Qwen3.5 architecture, using a mixture of experts design that wakes only part of the network.
The company said the system spent 16 days building a command line tool alone.
Qwen3.8-Max ranked as the top Chinese entry among text models on Arena.AI, though it still trails Claude Fable 5 and three Opus variants from Anthropic. On the platform leaderboard for models that read images and other visual material, the system placed second globally. Open weights follow next week, the first such release for a Max class Qwen model. Alibaba plans to publish them on Hugging Face and ModelScope.
Also Read: AI Infrastructure, Payday Lender's Desperate $1B Pivot To Dominate Data Centers
DeepSeek V4-Flash Undercuts Claude Fable 5
DeepSeek released V4-Flash on Friday, charging $0.14 per million input tokens and $0.28 per million output tokens. Artificial Analysis estimated its average cost at 3 cents per test, against 86 cents for Moonshot AI's Kimi K3, $1.86 for OpenAI's GPT-5.6 Sol and $3.15 for Claude Fable 5.
"Chinese AI companies have found an important market," said Lian Jye Su, chief analyst at research firm Omdia, who said many business workflows need only a system that is affordable and accessible.
Chinese AI Price War Reshapes Model Economics
Open weight releases let developers download the settings behind a model and adapt them, a route Anthropic, OpenAI and Google have not taken so far. V4-Flash scored 50 out of 100 on the Artificial Analysis Intelligence Index, a mark that ties Google's Gemini 3.6 Flash.
Kimi K3 reached 57, while flagship systems from Anthropic and OpenAI landed about nine points above that. The cheaper model also ran verbose, generating roughly twice the median token volume during the test.
DeepSeek drew global attention in early 2025, when its R1 and V3 models triggered a selloff in technology stocks and raised doubts about U.S. spending on AI. The company later made a 75% price cut permanent, and rivals including Moonshot, MiniMax, Z.AI, ByteDance and Alibaba have kept trimming prices since.
Read Next: Foldable iPhone Rumors Get Specific: $2,500, No Face ID, No Telephoto