Alibaba-backed AI company releases open-source model with per-task cost below OpenAI's flagship

An AI company in which Alibaba holds a 36% stake has released a new open-source model, claiming its per-task inference cost is lower than OpenAI's flagship offering. The news signals another front in the accelerating cost war between open-source and proprietary AI models.

The release adds to a growing trend started by Chinese open-source models like DeepSeek, which challenged GPT-4 with dramatically lower inference costs. Each new model in this lineage tightens the pressure on closed-source vendors to justify their pricing.

The company's choice of the open-source path aligns with Alibaba's broader AI strategy. Alibaba itself maintains the Tongyi Qianwen (Qwen) model family while investing across a portfolio of AI startups to build an ecosystem.

Per-task costs below OpenAI's flagship mean developers and enterprises face a lower economic barrier to accessing near-flagship AI capabilities through this open-source alternative. This is particularly attractive for small and medium businesses with tighter budgets.

However, cost alone does not guarantee market impact. The model's actual performance on reasoning, multi-turn conversation, code generation, and other critical tasks will determine whether it can truly compete with OpenAI's offerings.

The launch reflects a broader strategy among Chinese AI companies: combining open-source distribution with aggressive cost structures to challenge US AI leaders. The approach, validated by DeepSeek and others, is becoming a defining pattern in the global AI landscape.

Full technical reports and benchmark comparisons have not yet been published, and the industry is watching closely for performance data against OpenAI's flagship.