All NewsEducationTV
Equities & FundsCrypto & Digital AssetsAI & TechnologyBusiness & CorporateUS Politics & PolicyGeopolitics & Global RiskMacro, Rates & FXCommodities & EnergyEuropean Politics & MarketsAsia-PacificReal Estate & Property
Story archiveAll categories
← All Stories

Alibaba Releases Qwen3.8-Max AI Model Open-Source

Created at 4 Aug · 5:21 PM1 source↑ Market-relevant
IN SHORT

Alibaba has released its most capable AI model to date, Qwen3.8-Max, as open weights. The model boasts 2.4 trillion parameters and is designed for efficiency, allowing smaller businesses to utilize advanced AI without massive infrastructure costs. It also ships with setup instructions for rival tools from Anthropic and OpenAI.

✉Newsletter

PiQ Daily

Pick your topics. Get only what matters, on your cadence.

Key Numbers

2.4 trilliontotal parameters in Qwen3.8-Max
95 billionparameters switched on at any moment
16 daystime spent building a coding tool autonomously
2.7 pointsexceeded research paper results by
458 of 526human teams finished behind Qwen3.8-Max in a contest
nearly 30%lower intelligence cost compared to Claude Fable 5
15first-place spots for Fable 5 across 31 tests
9first-place spots for GPT-5.6 Sol across 31 tests
7first-place spots for Qwen3.8-Max across 31 tests
1coding test won by Qwen3.8-Max
61%share of tokens on OpenRouter for Chinese open-weight models by mid-2026

Who's Involved

Alibaba
released Qwen3.8-Max AI model as open weights
Anthropic
developer of Claude Code and Fable 5 AI models
OpenAI
developer of Codex and GPT-5.6 Sol AI models
Meta
developer of Llama AI models
Alibaba Releases Qwen3.8-Max AI Model Open-Source

↳ Why This Matters

Alibaba's decision to release its advanced Qwen3.8-Max AI model as open weights democratizes access to cutting-edge AI, potentially accelerating innovation for smaller businesses and researchers. It also signals a strategic shift in the competitive landscape of large language models, particularly concerning the growing influence of Chinese AI development amidst global regulatory scrutiny.

Key facts

  • Alibaba has released its Qwen3.8-Max AI model as open weights.
  • The model features 2.4 trillion parameters, with 95 billion active at any given time.
  • Qwen3.8-Max demonstrated its capabilities by autonomously developing a coding tool and reproducing research paper results.
  • It is designed for efficiency, making advanced AI accessible to smaller entities.
  • The model includes compatibility instructions for Anthropic's Claude Code and OpenAI's Codex.
  • While leading in multimodal tasks, Qwen3.8-Max performed less competitively in specific coding benchmarks compared to rivals.

Alibaba has released its Qwen3.8-Max large language model as open weights, a move that makes its most capable AI model publicly accessible. This release is significant as it marks the first time Alibaba has offered a Max-class Qwen model without charge, aiming to democratize access to state-of-the-art AI technology.

The Qwen3.8-Max model boasts a total of 2.4 trillion parameters, with 95 billion active at any given time, designed for efficiency and reduced resource requirements. This allows smaller businesses and research labs to run advanced AI models without the need for massive data center infrastructure. The model demonstrated its capabilities through extended autonomous tasks, including building a coding tool over 16 days and reproducing research paper results with an improvement of 2.7 points.

Notably, Qwen3.8-Max is designed to work within rival tools, shipping with setup instructions for Anthropic's Claude Code and OpenAI's Codex. While Alibaba's model leads in multimodal tasks like document and spatial reasoning, its performance in specific coding benchmarks was less dominant compared to competitors such as Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol. However, the model's cost efficiency is highlighted, estimated to be nearly 30% of Claude Fable 5's charges, making it a more economical option for complex tasks.

This open release represents a strategic shift for Alibaba, which had previously moved towards closed, paid models. The timing of the release coincides with a surge in the adoption of Chinese open-weight models and potential government restrictions on AI model exports. Alibaba's decision to make Qwen3.8-Max freely available positions it to gain distribution advantages, even if it does not top every benchmark.

Frequently asked questions

Qwen3.8-Max is a large language model developed by Alibaba, released as open weights. It features 2.4 trillion parameters and is designed for efficiency and accessibility.

It is Alibaba's first Max-class Qwen model released as open weights, making advanced AI technology freely available to a wider audience, including small businesses and labs.

While Qwen3.8-Max leads in multimodal tasks and offers high efficiency, it performed less competitively in specific coding benchmarks compared to models like Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol.

The release could accelerate AI adoption by smaller entities and signifies a strategic move by Alibaba in the competitive AI market, especially given potential global export controls on AI models.

What Happens Next

01Qwen3.8-Max weights will be available on Hugging Face and ModelScope next week.

Get the newsletter.

Pick the topics you actually care about. We'll email when there's news worth your time, on the cadence you choose. Cancel any time from your account.

Cadence

How It Developed

Alibaba released Qwen3.8-Max, its most capable AI model, as open weights.
The model has 2.4 trillion parameters, with 95 billion active at any time.
Qwen3.8-Max demonstrated endurance by building a coding tool autonomously over 16 days.
It reproduced a research paper's results, exceeding the original by 2.7 points.
The model includes setup instructions for Anthropic's Claude Code and OpenAI's Codex.
In coding tests, Qwen3.8-Max won one out of twelve, while rivals like Fable 5 and GPT-5.6 Sol performed better.
Qwen3.8-Max leads in multimodal tasks such as document and spatial reasoning.
The model's cost efficiency is estimated at nearly 30% of Claude Fable 5's charges.

Sources

T1
Alibaba Just Gave Away Its Best AI Model For Free, Almost Matching Claude and ChatGPTDecrypt

Related Stories

US AI Leaders Embrace Chinese Open-Weight Models, Questioning Safety Claims
4 Aug · 4:16 PM
China's MiniMax restricts global access to new AI video model due to copyright issues
4 Aug · 12:06 PM
Crypto firms seek access to advanced AI models for security
4 Aug · 6:16 AM
Chinese AI Firm Accused of Manipulating Ranking to Overtake Nvidia
4 Aug · 10:36 AM
CoreWeave expands into Indonesia with three new data centers
4 Aug · 1:08 PM