A change to the licensing of Meta’s next LLaMA model could widen the choices available to companies building chatbots. The Information reported that Meta plans to let other companies use and profit from a new version of its large language models. The plans were described by a person familiar with them.
That would mark a shift from the terms attached to the first LLaMA release, which permitted research use but barred commercial applications. A commercial license could make the model relevant to businesses that want to build products around language models without relying only on proprietary services.
From research release to commercial option
Meta released LLaMA in February to help researchers develop alternatives to proprietary models such as ChatGPT. The model family has since become a foundation for open-source chatbot projects, including Alpaca, Vicuna, and OpenAssistant.
Those projects show how a model can support work beyond the organization that created it. But research-only licensing puts a limit on that activity: a company may be able to study or experiment with a model, yet still be unable to use it in a commercial product. A license allowing commercial use could remove that particular barrier.
The reported plan concerns a new model, referred to as LLaMA v2. The source does not establish the precise terms of a future license, so the extent of the change remains uncertain. Still, permission to use and profit from the model would make it a different proposition for developers than the research-only release.
Why chatbot builders may pay attention
Companies developing chatbots often weigh the capabilities of a model alongside the terms under which they can use it. A commercially usable open-source model could offer another path for teams seeking alternatives to proprietary models from companies such as OpenAI and Google.
The article describes the existing LLaMA as on par with GPT-3, and says instruction tuning can bring it to the performance of GPT-3.5. A newer model could move closer to GPT-4, though that is presented as a possibility rather than a confirmed result.
If the next release combines useful performance with commercial access, it could attract developers who want to build on an open model. That might put pressure on proprietary providers by giving businesses another option. The scale of any shift would depend on the new model and its eventual license, neither of which is detailed in the report.
Meta could benefit from outside progress
Opening a model to commercial use could also serve Meta’s own interests. The source argues that a broader community of developers and researchers could contribute new ideas and advances that Meta might incorporate into AI models used in its services.
Meta already uses language models for content moderation and translation, and the company has plans for chatbots on Instagram. More activity around LLaMA could therefore connect outside experimentation with products and systems Meta is developing internally.
This creates a two-way dynamic: developers gain a model they can potentially use in products, while Meta may gain from the work that builds on it. Whether that exchange produces meaningful improvements will depend on what the company releases and how the community responds.
The license will shape the impact
The reported possibility matters because licensing can determine whether an AI model is only a research resource or a basis for commercial products. For chatbot developers, a change could expand the set of models they can consider. For the market, it could test whether open-source competition can challenge established proprietary services.
For now, the commercial availability of LLaMA v2 remains a reported plan, not a confirmed set of license terms. The source also notes that Zuckerberg recently discussed a possible LLaMA v2 in a podcast. The practical consequences will become clearer when Meta provides details about the model and how companies may use it.