Chinese language tech firm Alibaba on Monday launched Qwen3, a household of AI fashions the corporate claims matches and in some instances outperforms the very best fashions out there from Google and OpenAI.
A lot of the fashions are — or quickly might be — out there for obtain beneath an “open” license from AI dev platform Hugging Face and GitHub. They vary in measurement from 0.6 billion parameters to 235 billion parameters. Parameters roughly correspond to a mannequin’s problem-solving expertise, and fashions with extra parameters typically carry out higher than these with fewer parameters.
The rise of China-originated mannequin collection like Qwen have elevated the strain on American labs equivalent to OpenAI to ship extra succesful AI applied sciences. They’ve additionally led policymakers to implement restrictions geared toward limiting the power of Chinese language AI corporations to acquire the chips crucial to coach fashions.
Introducing Qwen3!
We launch and open-weight Qwen3, our newest massive language fashions, together with 2 MoE fashions and 6 dense fashions, starting from 0.6B to 235B. Our flagship mannequin, Qwen3-235B-A22B, achieves aggressive leads to benchmark evaluations of coding, math, normal… pic.twitter.com/JWZkJeHWhC
— Qwen (@Alibaba_Qwen) April 28, 2025
Based on Alibaba, Qwen3 fashions are “hybrid” fashions within the sense that they will take time and “purpose” by complicated issues or reply easier requests rapidly. Reasoning allows the fashions to successfully fact-check themselves, just like fashions like OpenAI’s o3, however at the price of greater latency.
“We’ve seamlessly built-in pondering and non-thinking modes, providing customers the pliability to regulate the pondering finances,” wrote the Qwen crew in a weblog publish. “This design allows customers to configure task-specific budgets with higher ease.”
A number of the fashions additionally undertake a mix of consultants (MoE) structure, which may be extra computationally environment friendly for answering queries. MoE breaks down duties into subtasks and delegates them to smaller, specialised “knowledgeable” fashions.
The Qwen3 fashions help 119 languages, Alibaba says, and have been skilled on a dataset of practically 36 trillion tokens. Tokens are the uncooked bits of knowledge {that a} mannequin processes; 1 million tokens is equal to about 750,000 phrases. Alibaba says that Qwen3 was skilled on a mix of textbooks, “question-answer pairs,” code snippets, AI-generated knowledge, and extra.
These enhancements, together with others, significantly boosted Qwen3’s capabilities in comparison with its predecessor, Qwen2, says Alibaba. Not one of the Qwen3 fashions are head and shoulders above top-of-the-line current fashions like OpenAI’s o3 and o4-mini, however they’re sturdy performers nonetheless.
On Codeforces, a platform for programming contests, the most important Qwen3 mannequin — Qwen-3-235B-A22B — simply beats out OpenAI’s o3-mini and Google’s Gemini 2.5 Professional. Qwen-3-235B-A22B additionally bests o3-mini on the newest model of AIME, a difficult math benchmark, and BFCL, a take a look at for assessing a mannequin’s means to “purpose” about issues.
However Qwen-3-235B-A22B isn’t publicly out there — no less than not but.
The most important public Qwen3 mannequin, Qwen3-32B, continues to be aggressive with quite a lot of proprietary and open AI fashions, together with Chinese language AI lab DeepSeek’s R1. Qwen3-32B surpasses OpenAI’s o1 mannequin on a number of assessments, together with the coding benchmark LiveCodeBench.
Alibaba says Qwen3 “excels” in tool-calling capabilities in addition to following directions and copying particular knowledge codecs. Along with the fashions for obtain, Qwen3 is obtainable from cloud suppliers, together with Fireworks AI and Hyperbolic.
Tuhin Srivastava, co-founder and CEO of AI cloud host Baseten, mentioned that Qwen3 is one other level within the development line of open fashions protecting tempo with closed supply techniques equivalent to OpenAI’s.
“The U.S. is doubling down on limiting gross sales of chips to China and purchases from China, however fashions like Qwen 3 which might be state-of-the-art and open … will undoubtedly be used domestically,” he instructed TechCrunch. “It displays the fact that companies are each constructing their very own instruments [as well as] shopping for off the shelf through closed-model corporations like Anthropic and OpenAI.”
