OpenAI Shares Benchmark Scores of o3 Series AI Models, Offers Unlimited Access of Sora to Paid Subscribers

OpenAI shared the benchmark scores of the o3 sequence of synthetic intelligence (AI) fashions on Friday. As successor to the lately launched o1 sequence of AI fashions, the corporate claimed that o3 considerably surpasses the capabilities of the older model. The sequence has two fashions, the o3 and o3-mini, and is targeted on superior reasoning-based duties. Currently, the AI fashions are being made out there for public security testing, and the o3-mini mannequin will probably be launched within the public area early subsequent yr. Notably, OpenAI’s announcement got here simply days after Google launched its Gemini 2.0 Flash Thinking Mode AI mannequin.

OpenAI Shares Benchmark Scores of the o3 Series AI Models

On day 12 of the AI agency’s 12-day delivery schedule, OpenAI announced the o3 sequence, the successor to the o1 AI mannequin. During a reside stream, CEO Sam Altman highlighted that the title o2 was dropped as a result of UK telecommunications service supplier Telefonica which makes use of the identical model title.

OpenAI highlighted that regardless of the announcement, the fashions will at present not be out there within the public area. The firm has began offering each o3 and o3-mini in early entry to chose exterior researchers for public security testing. Those fascinated by collaborating within the course of can apply here. The utility course of ends on January 10.

The firm highlighted that each the o3 sequence AI fashions will probably be extra highly effective than their predecessors, and can be capable to carry out extra advanced duties in coding, arithmetic, and pure language processing. However, its full vary of capabilities was not revealed.

Additionally, OpenAI shared some benchmark evaluations of the o3 AI fashions throughout inner testing. The firm claimed that o3 scored 71.7 p.c within the SWE-bench benchmark and 96.7 p.c within the AIME 2024 benchmark, surpassing o1. However, a full benchmark analysis will solely be doable as soon as the fashions can be found within the public area. The o3-mini is predicted to be launched in January 2025.

OpenAI Offers Unlimited Access of Sora to Paid Subscribers

Separately, as a “day-13 bonus”, Altman introduced in an X (previously referred to as Twitter) post that the ChatGPT Plus and Teams subscribers will get limitless entry to Sora over the last two weeks of December. The CEO highlighted that this was doable as places of work shut in the course of the Christmas interval leading to diminished server hundreds.

Rohan Sahai, a product lead at OpenAI added to the submit and stated that the corporate had upgraded Sora’s mix function and enabled shared hyperlinks to let customers share the AI-generated movies with mates, even when they don’t have an account.