Newshot
Account

Baidu's 李彦宏 launches "Token Exchange," a marketplace for large models: claims neutrality and trustworthiness, rates model providers

AI summary

At Baidu World 2026 on November 12, Baidu founder 李彦宏 announced a new large-model service platform called Token Exchange, described as a neutral, trustworthy marketplace for large models that he likened to a stock exchange. The platform claims to run full admission testing of model resources — including model and fingerprint comparison, capability tests, performance benchmarks and availability tests — plus continuous monitoring of quality parameters, and to score providers on service quality, delivery, security compliance and commercial fulfillment with credit rating labels. One month after launch, it has more than 100 supply resources and over 700 billion average daily Token calls.

Translation(model translation, not the source's own text)

Baidu's 李彦宏 launches "Token Exchange," a marketplace for large models: claims neutrality and trustworthiness, rates model providers

IT Home, November 12, on-site report: At today's Baidu World Conference 2026 event, Baidu founder 李彦宏 announced the launch of a brand-new large-model service platform — Token Exchange.

李彦宏 said that when a developer or an enterprise truly wants to put models to use, they often find that the more models there are, the harder it is to choose. The same model may be expensive from one source and cheap from another, sources are all over the place, and quality varies widely; after all the choosing, they are still uneasy and afraid to use it.

Token Exchange was created to solve this problem, and is billed as a neutral and trustworthy large-model trading platform. 李彦宏 introduced that this exchange can be compared to a stock exchange. A stock exchange attracts good companies to list, letting investors discover and invest in good companies; Token Exchange, by contrast, hopes to bring together good models and model providers, so that enterprises and individuals can conveniently find, purchase, and use trustworthy models.

IT Home learned at the conference site that Token Exchange claims to conduct comprehensive admission testing of model resources, covering models, fingerprint comparison, model capability testing, performance metric testing, availability testing, and more, and will also continuously monitor model quality parameters. Through these means, Token Exchange aims to keep practices such as swapping in different models and falsely labeling parameters out the door as much as possible, ensuring that every model resource that enters is guaranteed.

Token Exchange will also score model providers across multiple dimensions such as service quality, fulfillment, security and compliance, and commercial performance, and issue credit rating labels. For model providers whose credit declines, Token Exchange will proactively remind users at the first opportunity. All of this data will be consolidated into publicly searchable credit records.

One month after Token Exchange went online, the number of supply resources has exceeded 100, and the average daily number of tokens called has surpassed 700 billion.

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by IT之家
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗