OpenAI’s ‘open’ AI Model to Utilize Cloud Assistance
OpenAI is preparing to unveil an AI system that will be freely available for download, marking a significant shift from their usual API-gated models. This open AI model, set for release in early summer, aims to outperform comparable offerings from Meta and DeepSeek.
The new AI model is expected to feature enhanced performance capabilities. Additionally, OpenAI might introduce a unique functionality that allows the model to connect to OpenAI’s cloud-hosted systems for better handling complex queries, increasing its competitiveness.
According to insiders, OpenAI executives have discussed enabling the open model to interface with cloud-hosted models during a meeting with developers. OpenAI CEO Sam Altman referred to this potential feature as a “handoff,” indicating that the open model could leverage the computational power of larger, more advanced models via the OpenAI API.
TechCrunch’s inquiries for comments from OpenAI went unanswered.
Should this handoff feature be implemented, the open model could access the OpenAI API, thereby utilizing more sophisticated models for enhanced computational performance. It remains uncertain whether the open model will have access to various tools available to OpenAI’s models, such as web search and image generation capabilities.
The concept of the handoff feature reportedly originated from a developer during one of OpenAI’s forums, gaining traction within the company. OpenAI has been actively engaging with developers through community feedback events to refine their upcoming model.
This hybrid approach, combining local models with powerful cloud systems, is reminiscent of Apple Intelligence, which utilizes a mix of device-specific models and those hosted in private data centers. For OpenAI, this strategy could not only drive incremental revenue but also integrate more open source developers into their premium ecosystem.
However, several aspects of how the handoff feature will function remain uncertain, including pricing and usage limits. The model is still in development, and certain features might be modified or may not materialize.
Sources indicate that OpenAI is developing a new model from the ground up for this release, rather than modifying an existing one. Although it may not match the performance of OpenAI’s o3 model, it is expected to surpass DeepSeek’s R1 reasoning model in specific benchmarks.
(Source: TechCrunch)
