OpenAI CEO Sam Altman interviewed at Bloomberg House on the opening day of the World Economic Forum (WEF) in Davos, Switzerland on Tuesday, January 16, 2024.
Chris Ratcliffe | Bloomberg | Getty Images
OpenAI announced a preview of the research for GPT-4.5, a general-purpose large language model. Initially, it will be open to software developers and Chatgpt Pro subscriptions.
Openai is in Blog Posts.
“Early testing showed that interaction with GPT −4.5 felt more natural. Its broader knowledge base, improved ability to follow user intentions, and a larger ‘EQ’ make it useful for tasks such as improving writing, programming, and solving practical problems,” Openai wrote. GPT-4.5 showed a lower hallucination rate than OpenAI’s GPT-4O and O1 models in one test. Report With Thursday’s release.
Meanwhile, OpenAI co-founder and CEO Sam Altman said Thursday that the new model will not be the most advanced of the benchmark X Postsdescribes the model as “a magic I’ve never felt before.”
Still, this version represents the advancement of one of the most popular startups in the world. Last month, CNBC reported that Openai was discussing A’s $40 billion funding. Valuation is $340 billion.
Finish 400 million Alex Paino, the company’s head of research, is a member of the company’s technician, said Openai intends to bring GPT-4.5 to people with Chatgpt Plus and team subscriptions. live streaming.
Paino said Chatgpt Edu and Enterprise users will gain access next week.
Altman wrote in a post on Thursday X that the GPT-4.5 is a “giant, expensive model.” “We really want to boot it to the Plus and Pro at the same time, but we’ve grown a lot and out of the GPU. We’re going to add thousands of GPUs next week and then roll it out to the Plus Tier.”
GPT-4.5 can be passed in preview Microsoft’s Azure AI Foundry, CEO Satya Nadella Announce Investing in X. Microsoft More than $13 billion In Openai, the startup model is included in its products. Microsoft also helps provide computing resources to OpenAI.
The AI ​​model was released a month after the market estimated the effective method adopted by Chinese laboratory DeepSeek. Market value NvidiaIts chips are widely used in AI model developers, rejected Nearly $600 billion Within one day.
Two weeks later, Ultraman was in a book X Posts The company hopes to better share the roadmap publicly.
Artman said the GPT 4-5 will appear, followed by the GPT-5, which will attract more OpenAI technology. The company also provides the so-called Inference Model This does a lot of calculations when people submit queries. In contrast, GPT-4.5 will be the company’s “last non-chain model”.
watch: The Future of Distillation and SoftBank, OpenAI Partnership
