In April this year,OpenAI Publicly acknowledged strategic mistakes in open-weight language models, and revealed plans to launch something new. AI As for the open-source model, at the time OpenAI CEO Sam Altman only mentioned it would come within the coming months. It appears the internal plan had tentatively set a June release. Earlier, Sam Altman provided an update on X, saying their open-source model would be delayed and is expected to arrive in the summer, emphasizing that it will be worth the wait.
OpenAI’s open-source model is confirmed to launch in the summer, so it won’t be coming in June.
Earlier, Sam Altman said on the social platform X: “We will spend a bit more time polishing our open-source model. That is, it is expected to launch later this summer, but not in June. Our research team has produced some unexpected and very impressive results, which we believe are very, very worth waiting for, but it will still need some time.”
we are going to take a little more time with our open-weights model, i.e. expect it later this summer but not june.
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.
— Sam Altman (@sama) June 10, 2025
OpenAI originally planned to launch a new open-source model in early summer, expected to have “reasoning” capabilities comparable to its own o-series models, as well as performance surpassing other open-source reasoning models such as DeepSeek’s R1. Additionally, OpenAI is also considering adding multiple complex features to this open-source AI model to enhance its competitiveness.
However, since OpenAI made this information public, various companies have successively released stronger new models, such as Mistral’s first AI reasoning model “Magistral,” and Qwen, which released the Qwen 3 series of hybrid AI reasoning models at the end of April, including Qwen3-235B-A22B and Qwen3-30B-A3B. DeepSeek is no exception, launching a new DeepSeek-R1-0528 on May 28, with stronger capabilities in mathematics, coding, and logical reasoning, while also reducing the likelihood of hallucinations.
Therefore, this may be one of the reasons OpenAI delayed its launch.
Image source:Medium
According to foreign media TechCrunch Earlier reports said that OpenAI executives had discussed giving this open-source model the ability to connect to the company’s cloud AI model to handle more complex queries. However, it’s uncertain whether that will actually be added to the final version.
In any case, open-source models from OpenAI should hopefully not disappoint, especially since Sam Altman also emphasized “unexpected and very surprising results.” With OpenAI entering the fray, open-source models are likely to experience a period of rapid growth, which is quite exciting to look forward to.
OpenAI released the GPT-2 open-source model as early as 2019, but from then until the end of 2024, the company focused on commercial models, leaving no developments in its open-source efforts. After China’s DeepSeek burst onto the scene and became hugely popular early this year, OpenAI publicly admitted its previous strategic mistakes and expressed a willingness to put more effort into open-source models.
Besides OpenAI, at this WWDC 2025 event, Apple also introduced Foundation Models Framework and 3B On-device Model：
Image source:Apple
Source: KOCPC Chinese