TMTPOST -- OpenAI   is escalating the   price   war   in   large language   model   ( LLM )   while seeking   partnership   with a   prominent   artificial   intelligence ( AI )   rival   to meed its   compute needs amid   fierce   rivaly.

Credit：Alibaba Cloud

OpenAI CEO Sam Altman on   Tuesday   announced the ChatGPT   developer cut costs   for   both   input   and output tokens   of   o3   by 80%. OpenAI   released   o3 in April, touting its as   "our most   powerful   reasoning model   that   pushes the   frontier   across   coding, math, science, visual perception, and more."   The company said O3 performs especially strongly at visual tasks like analyzing images, charts, and graphics, and   in evaluations by external experts,the model   makes 20%   fewer major errors than OpenAI o1 on difficult, real-world tasks — especially excelling in areas like programming, business/consulting, and creative ideation.

In a   post on social media X   on Tuesday, Altman also showed upbeat on   o3-pro, an   enhanced   variant   of o3, saying "think you ’ ll also be happy with o3-pro   pricing for the performance."   O3-pro is priced at $20 per million input tokens and $80 per million output tokens in the API. In comparison, the cost of using   o3 is now $2 per million input tokens and $8 per million output tokens, with an extra discount of $0.50 per million tokens when the user enters information that has been "cached," or is stored and identical to what they provided before.

"Like o1-pro, o3-pro is a version of our most intelligent model, o3, designed to think longer and provide the most reliable responses,"   OpenAI   said in a blog.   O3-pro is available in the model picker for Pro and Team users starting Tuesday, replacing o1-pro,   and Enterprise and Edu users will get access the week after,   according to OpenAI.   O3-pro is also live in   OpenAI ’ s developer   API on   Tuesday.

In expert evaluations, reviewers consistently prefer o3-pro over o3 in every tested category and especially in key domains like science, education, programming, business, and writing help. Reviewers rated o3-pro consistently higher for clarity, comprehensiveness, instruction-following, and accuracy, and academic evaluations show that o3-pro consistently outperforms both o1-pro and o3, per OpenAI.

OpenAI   was   finalized a   deal in May to tap   Google cloud   service   for   its   growing compute needs,   Reuters reported   earlier Tuesday. Under   the deal, Google ’ s   cloud   unit will supply new   computing capacity to OpenAI ’ s existing infrastructure   for training and running its models,   the report cited sources.

The unprecedented   partnership   with Google highlighted   how the   massive computing   demands for   training   and deployment of models   is reshaping   the   competitive dynamics   in AI. It is also part of OpenAI ’ s efforts to reduce dependence on   its top   investor   and key partner   Microsoft Corporation.   The deal has been reportedly under discussion   for a few months and was"previously blocked from signing a deal due to OpenAI ’ s lock-in with Microsoft."  

Microsoft, with its Azure cloud   service, has had exclusively   offered   computing capacity   for   OpenAI   until January.   The   software giant on   January 21   said it was   no longer OpenAI ’ s exclusive cloud provider, though it   now   has the   "right of first refusal"   before   OpenAI   checks with other parties. OpenAI in January   took a marjor step   toward diversifying   its compute sources   with annoucement of the Stargate   Project,   a joint venture with OpenAI, Oracle and Softbank to invest billions of dollars in AI infrastructure in the U.S.

OpenAI was reported in February to anticipate   a   major shift in the   next five years around whom it gets   most of   its   computing power from. The company expected   to get three quarters of   its data center   capacity from Stargate   by 2030,   The Information   reported.   That would represent a reversal from today and the next few years, when its power and data center capacity is mostly handled by Microsoft.