OpenAI Delays Open-Weight Model Release, Promises "Unexpected and Amazing" Advancements

OpenAI's open model is delayed

The release of OpenAI's first open model in years will be delayed until later this summer, CEO Sam Altman announced in a post on X on Tuesday. Altman said the open model would be released sometime after June. "We are going to take a little more time with our open-weights model, i.e. expect it later this summer but not June," he wrote. "Our research team did something unexpected and quite amazing and we think it will be very very worth the wait, but needs a bit longer." OpenAI was targeting an early summer release date for its open model, which is slated to have similar "reasoning" capabilities to OpenAI's o-series of models. OpenAI aims for its open model to top the performance of other open reasoning models, such as DeepSeek's R1. In the months since OpenAI first announced its intent to release an open model, the space has become more competitive. On Tuesday, Mistral -- another AI lab that often releases open models -- released its first family of AI reasoning models, called Magistral. In April, the Chinese AI lab Qwen released a family of hybrid AI reasoning models that can switch off between taking time to "reason" through problems and also giving traditional, quick responses. Beyond increasing its performance on benchmarks, OpenAI has also considered adding several complex features to its open AI model to make it more competitive. TechCrunch previously reported that OpenAI leaders have discussed enabling the open AI model to connect to the company's cloud-hosted AI models for complex queries. However, it's unclear if these features will make it into the final open model. The release of OpenAI's open model seems to be important for the company's relationship with researchers and developers. Altman has previously said that OpenAI has landed on the "wrong side of history" when it comes to open sourcing its models. To rectify that image, the company faces immense pressure to release an open model that is competitive with the industry's best open offerings.

OpenAI is delaying its next model -- but promises something amazing

OpenAI is no stranger to messy timelines. The company has a history of delays, changes and strange turns in the road. But with its latest delay, the extended timeline might actually be good news. "We are going to take a little more time with our open-weights model, i.e. expect it later this summer but not June," Sam Altman, CEO of OpenAI announced in a post on X. "Our research team did something unexpected and quite amazing and we think it will be very very worth the wait, but needs a bit longer." OpenAI has been teasing news of a new open-weight model for a while now, originally planning for an early summer release. This would offer reasoning capabilities similar to OpenAI's o-series models, but open to the public. Openweight models are a type of AI model where the trained parameters are publicly available. In other words, they are fully trained AI systems that any individual or company can download and use for their own projects without having to perform the intensive process of building and training a model. For now, Altman hasn't stated why the model is being pushed back. All we know is that the team has apparently done something "unexpected and quite amazing". The AI market, especially the open-weight one, has become a lot more competitive in recent months. New competitors are coming in and doing it cheaper than OpenAI. Mistral, another big AI lab, released its first range of AI reasoning models and competitors like Deepseek, Microsoft and Google are all also deep into the open-weight AI market. If the OpenAI team discovered something they could add to the model to beef it up against the competition, a short delay could be well worth it for the company to get an edge. For most people, all eyes will be on the release of GPT-5 -- the latest update to OpenAI's ChatGPT model. While no official date has been given for this, most analysts believe it will be launching in the next couple of months, likely in July. The original plan was May, which we have obviously passed. Now almost halfway through June, July is looking quite likely. Back in February this year, Altman posted on X outlining a roadmap for future updates. At this time, it was suggested that GPT-5 would be made available to all ChatGPT users, including those on a free plan. Responding to a comment, Altman also said the model would be here in "weeks/months". As we mentioned above though, OpenAI is known for its delays. For now, we can all simply hope for a release date in the near future.

OpenAI delays open-weight model launch to later this summer

OpenAI CEO Sam Altman announced Tuesday via X that the company's first open model release in years is delayed until later this summer but sometime after June. The delay accommodates unexpected advances by OpenAI's research team. Altman stated on X, "[W]e are going to take a little more time with our open-weights model, i.e. expect it later this summer but not [J]une. [O]ur research team did something unexpected and quite amazing and we think it will be very very worth the wait, but needs a bit longer." OpenAI initially planned for an early summer release, aiming for "reasoning" capabilities akin to its o-series models, seeking to surpass models like DeepSeek's R1. Since OpenAI's announcement, the field has grown increasingly competitive. Mistral, another AI lab, recently launched its Magistral family of AI reasoning models. In April, Qwen, a Chinese AI lab, introduced hybrid AI reasoning models capable of alternating between deliberate "reasoning" and quick responses. Beyond benchmark performance, OpenAI has considered integrating complex features into its open AI model, including potential connectivity with the company's cloud-hosted AI models for complex queries. The open model's release is significant for OpenAI's relations with researchers and developers. Altman has acknowledged OpenAI's past stance on open sourcing as being on the "wrong side of history." The company now faces pressure to deliver a competitive open model.