OpenAI postpones open model launch once more
▼ Summary
– OpenAI CEO Sam Altman announced an indefinite delay in releasing the company’s open model to conduct additional safety testing and review high-risk areas.
– The open model, highly anticipated alongside GPT-5, will allow developers to freely download and run it locally, aiming to reinforce OpenAI’s leadership in AI.
– Developers must wait longer for OpenAI’s first open model in years, expected to have reasoning capabilities similar to its o-series models and be best-in-class.
– The open AI model ecosystem grew more competitive with Chinese startup Moonshot AI launching Kimi K2, which outperforms OpenAI’s GPT-4.1 on certain benchmarks.
– OpenAI’s VP of research, Aidan Clark, stated the model is “phenomenal” but requires more time to meet the company’s high standards for open-source release.
OpenAI has once again delayed the launch of its highly anticipated open AI model, marking the second postponement in recent months. Originally slated for release next week, the company now says the model requires further safety evaluations before being made available to developers.
In a recent post on X, OpenAI CEO Sam Altman explained the decision, emphasizing the irreversible nature of releasing model weights. “We need more time for safety testing and reviewing high-risk areas,” he wrote. “Once the weights are out, they can’t be undone. This is uncharted territory for us, and we want to ensure we get it right.”
The delay comes as OpenAI seeks to reinforce its position as a leader in AI development amid intensifying competition. Rivals like xAI, Google DeepMind, and Anthropic are pouring billions into their own AI initiatives, making the stakes higher than ever. The open model, once released, will allow developers to download and run it locally, a stark contrast to OpenAI’s cloud-based offerings like GPT-5, which is also expected soon.
Reports suggest the open model will boast reasoning capabilities comparable to OpenAI’s proprietary models, potentially setting a new benchmark for open-source AI. However, the postponement means developers must wait longer to test its capabilities firsthand.
Meanwhile, the open AI landscape is heating up. Just this week, Chinese startup Moonshot AI unveiled Kimi K2, a trillion-parameter model that reportedly outperforms OpenAI’s GPT-4.1 in certain coding benchmarks. This development adds pressure on OpenAI to deliver a model that stands out in an increasingly crowded field.
Earlier this summer, Altman hinted at an “unexpected and quite amazing” breakthrough related to the open model but provided no further details. Aidan Clark, OpenAI’s VP of Research, echoed this sentiment, stating, “We believe the model is phenomenal in capability, but our standards for open-source releases are exceptionally high. We need more time to ensure it meets our expectations across all dimensions.”
While OpenAI has explored integrating the open model with its cloud-based AI services for advanced queries, it remains uncertain whether this functionality will be included in the final release. For now, the AI community will have to wait as OpenAI takes extra precautions before unveiling its latest innovation.
(Source: TechCrunch)
