STOCK TITAN

Baidu Unveils Latest ERNIE Models Amid Broader Full-Stack AI Lineup, and Debuts Token Exchange to Power Intelligent Organizations

Token Exchange has onboarded more than 100 model resources, while Baidu has begun shipping its M100 inference chip.

(Neutral)

Sentiment and the balance of points

Rhea-AI Sentiment reads the wording of the document, how positive or negative its language is on a 1 to 5 scale. The balance of points shown with the takes weighs what the document actually discloses, so the two can disagree, for example when a trial that missed its main goal is described in upbeat language.

Tags
AI

Baidu (BIDU) unveiled new AI models and launched Token Exchange, a marketplace connecting model service providers with customers. ERNIE 5.5 Flash Preview is available through its Qianfan platform at RMB 1 per million input tokens and RMB 4 per million output tokens; model weights are scheduled for release by November 25.

One month in, Token Exchange handles a daily average of more than 50 million calls and over 700 billion tokens. Baidu also upgraded its agents and introduced the Tianchi 1024-card Supernode. Its Kunlunxin M100 inference chip has shipped, while the company expects M300 shipments in 2027. Apollo Go has launched fully driverless commercial operations in Abu Dhabi and Dubai.

Loading...
Loading translation...
7 points · 0 major

How this balance works

Rhea-AI gives every point it takes from this document a weight. Minor counts 1, Moderate 3 and Major 9, so one Major point outweighs several Minor ones. The bar adds up the weights on each side, and when neither side holds more than 65% of the total the balance reads Mixed.

It reads the document as published, with the same rules for every company, and it does not look at what the market expected or at how the stock traded, so a point can be objectively good on a day the stock falls.

Rhea-AI Sentiment measures something else, the tone of the wording.

0 major · 0 points

Hollow bars mark forward-looking points. How the balance works

Positive

  • Moderate pointToken Exchange handles daily averages exceeding 50 million calls and 700 billion tokens, one month in.
  • Moderate pointApollo Go launched fully driverless commercial operations in Abu Dhabi and Dubai.
  • Minor pointERNIE 5.5 Flash Preview API launched at RMB 1 per million input tokens and RMB 4 per million output tokens.
  • Minor pointDaily active agents inside Baidu rose from 5,000 to 37,000 over the past year, more than 600%.
  • Minor pointInternal agents now complete more than 1 million tasks weekly, up from 84,000.
2 minor points
  • Minor pointKunlunxin M100 inference chip has shipped as planned.
  • Minor point. Forward-looking: it has not happened yet and may not happen.Kunlunxin M300 is expected to ship in 2027.

Negative

  • None.

News Explained

Baidu reports Apollo Go covers 28 cities worldwide and has logged more than 350 million autonomous kilometers, including over 240 million in fully driverless mode, placing the newly announced Abu Dhabi and Dubai commercial launches within a broader existing deployment.

Key Figures

Total model parameters: 280 billion parameters Context window: 1M Input API price: RMB 1 per million input tokens +5 more
Total model parameters
280 billion parameters
ERNIE 5.5 Flash Preview
Context window
1M
ERNIE 5.5 Flash Preview
Input API price
RMB 1 per million input tokens
ERNIE 5.5 Flash Preview
Output API price
RMB 4 per million output tokens
ERNIE 5.5 Flash Preview
Cached input API price
RMB 0.04 per million tokens
ERNIE 5.5 Flash Preview
Model resources onboarded
More than 100
Token Exchange, one month after launch
Daily average calls
More than 50 million
Token Exchange
Daily token volume
Over 700 billion tokens
Token Exchange

Previous AI Reports

2 past events · Latest: Nov 13
Same Type 2 events
  1. Nov 13

    AI model launch

    24h Move
    -6.4%

    Introduced ERNIE 5.0 and related AI applications in a prior foundation-model lineup launch.

  2. Apr 25

    AI model launch

    24h Move
    +1.4%

    Launched ERNIE 4.5 Turbo and X1 Turbo with API pricing and cost reductions.

24h Move is the share-price change in the day after each event; other market factors may also have contributed.

Key Terms

inference, context window, recursive self-improvement, all-optical interconnect
4 terms
inference technical
"foundation model designed for efficient inference"
Inference is the process of drawing a conclusion from available evidence or data, like a detective piecing together clues to form a likely story. For investors it matters because these judgments turn raw reports, test results, or market signals into expectations about future performance, risk, or regulatory outcomes—so how someone infers from the same facts can change investment decisions and valuation.
context window technical
"the model supports a 1M context window"
The context window is the amount of text an AI system can read and consider at one time — like the size of a desk where it lays out documents to make sense of them. For investors, a larger context window means the AI can analyze longer filings, full earnings calls or combined news streams without losing earlier details, improving the accuracy of summaries, risk signals and investment decisions.
recursive self-improvement technical
"proprietary recursive self-improvement (RSI) technology"
A process in which a software system or artificial intelligence makes changes to its own code, model or design in order to improve its performance or capabilities without direct human reprogramming. It matters to investors because it can rapidly accelerate a company's product development, competitive edge and potential value—like a machine that upgrades itself—while also changing the speed and uncertainty of technical progress and the regulatory and safety risks a business faces.
all-optical interconnect technical
"an all-optical interconnect supernode built for high-speed data transmission"
A communication link that carries data entirely as light between components or circuits without converting signals to electricity; it uses optical waveguides, lasers, modulators and photodetectors so photons transmit and receive information end-to-end. All-optical interconnects eliminate intermediate electrical–optical–electrical conversions, which can reduce latency and electrical power dissipation but requires integrated photonic hardware and precise alignment and signal management to work across chip, board or rack scales.

AI-generated analysis. How Rhea-AI works. Not financial advice.

See more from StockTitan in Google Search and AI answers. Adds StockTitan as a preferred source · opens Google
Add on Google
  • Robin Li, Co-founder and CEO of Baidu, outlined his vision for the next chapter of AI, centered on the intelligent organization, and introduced three principles of organizational evolution.
  • Baidu also introduced ERNIE 5.5 Flash Preview, the latest open-weight foundation model designed for efficient inference, and two application models ERNIE Wizshot Micro-Expression and ERNIE Light Deep Thinking.
  • Baidu launched Token Exchange, a neutral, transparent, and trusted marketplace connecting model service providers to customers.
  • Baidu also upgraded its agent lineup, including digital human agent Baidu Yijing, the general-purpose agent DuMate, document and creative workspace KOOKO, and self-evolving agent Famou 3.0.

BEIJING, Nov. 12, 2026 /PRNewswire/ -- Baidu, Inc. (NASDAQ: BIDU and HKEX: 9888), a leading AI company with strong internet foundation, today unveiled new offerings and upgrades across its full-stack AI portfolio spanning chips, cloud, models, and agents, designed to help enterprises transform into intelligent organizations. The new lineup includes Token Exchange, a trusted marketplace for model services; upgrades to its agents DuMate, KOOKO, Baidu Yijing, and Famou; ERNIE 5.5 Flash Preview, its latest foundation model, which balances intelligence with inference efficiency; as well as the new Tianchi 1024-card Supernode and Kunlunxin roadmap.

Robin Li, Co-founder and CEO of Baidu, speaks at Baidu World 2026

In his opening keynote, Robin Li, Co-founder and CEO of Baidu, described the intelligent organization as the next chapter of AI. Li explained why turning that technological progress into productivity gains means moving beyond traditional hierarchies. Just as AI needs energy to run, making full use of its capabilities requires an AI-native approach to how organizations operate. "If AI is the new electricity, then the intelligent organization is the new factory of the AI era, where productivity truly happens," Li said.

Li also laid out three principles for how organizations should evolve. The first is to break down job boundaries. With AI now enabling one person to act in multiple roles, Li said organizations will be built around tasks rather than jobs and become a network of tasks. "Jobs won't disappear, but they will converge into roughly two types: Builders, who create new things, and Sellers, who connect with people to forge emotional bonds and earn their trust," Li said. The second is to capture tacit knowledge and turn it into a shared company asset. "This is a capability that competitors can't copy, hire away, or buy. That's what a real moat looks like." The third is to design closed-loop processes. Li put forward a formula: measurability plus looping equals continuous evolution, so that "the organization becomes a system that never stops learning and evolving."

Robin Li, Co-founder and CEO of Baidu, explained the three principles for organizational evolution

Early results at Baidu have already validated the three principles. Over the past year, the average number of daily active agents (DAA) inside Baidu rose from 5,000 to 37,000, an increase of more than 600%. Agents now complete more than 1 million tasks a week, up from 84,000. Baidu has also built 4,300 workspaces and captured more than 20,000 organizational skills. According to Li, by turning practical experience into reusable skills, organizations can build shared capabilities that benefit the entire workforce.

Foundation and Application Models Advance in Parallel: ERNIE 5.5 Flash Preview, ERNIE Wizshot Micro-Expression and ERNIE Light Deep Thinking

Baidu also released its latest-gen foundation model ERNIE 5.5 Flash Preview. With 280 billion total parameters and 14 billion parameters activated per token, the model supports a 1M context window and performs strongly in coding, 3D modeling, knowledge work, and content creation.

ERNIE 5.5 Flash Preview balances intelligence with inference efficiency. Across multiple benchmarks, including Terminal-Bench and AutomationBench, it achieved competitive scores while maintaining extremely fast decoding speed. The model API is now available on Baidu AI Cloud's Qianfan platform, and the model weights will be open-sourced by November 25. API pricing is RMB 1 per million input tokens and RMB 4 per million output tokens, with cached input at just RMB 0.04 per million tokens. The model is also available for free on OpenRouter. ERNIE 5.5 Flash Preview is optimized with ERNIE Evolve, Baidu's proprietary recursive self-improvement (RSI) technology, to boost inference efficiency and reduce cost.

Baidu also introduced two application models, the ERNIE Wizshot Micro-Expression model and the ERNIE Light Deep Thinking model, available via API on Baidu AI Cloud's Qianfan platform. The ERNIE Wizshot Micro-Expression specializes in highly expressive video creation. It processes text, images, and speech together and can support long-form text input of up to 4,000 Chinese characters. It leads the industry for podcasts and short animated dramas at half the typical cost. Originally developed for multimedia search, the ERNIE Light Deep Thinking is designed to handle complex requests and tool calls at search-scale speed. It pairs a fast-thinking model that serves users with a slow-thinking model that continuously refines it in a closed loop. It also incorporates Baidu Sparse Attention (BSA), a proprietary technology that compresses historical memory during decoding to reduce inference costs without compromising model performance.

Token Exchange: A Trusted Marketplace for Model Services

Looking at the broader market, demand for model services is growing fast. To make high-quality model services easier to find and use, Baidu launched Token Exchange, a neutral, transparent, and trusted marketplace that connects AI model providers with customers. Customers can compare model services side by side, with clear pricing and performance data, and access them in one place. Quality providers, in turn, gain a channel to access more customers. Li described it as an exchange for foundation models, where businesses and individuals can easily find, buy, and use trusted model tokens.

Token Exchange rests on two pillars of trust: verification and credit rating. To verify each resource, it compares model fingerprints and tests capability, performance, and availability. To rate each provider, it runs regular spot checks and monitors calls end to end in real time. One month in, Token Exchange has onboarded more than 100 model resources and now handles a daily average of more than 50 million calls and over 700 billion tokens. 

Over time, Token Exchange is expected to help put underused compute to work and make the AI and foundation model ecosystem healthier and more efficient.

Baidu Yijing, DuMate, KOOKO, and Famou: Agent Upgrades Put the Intelligent Organization to Work

At the event, Baidu Yijing, the hyper-realistic digital human agent, introduced short animated drama creation covering concept, script, storyboard, and final cut, turning a single idea into a finished production in one stop. With one-click localization, creators can then take their dramas to global audiences. Built on Baidu's proprietary video generation model, production costs are just RMB 0.2 per second of video.

Baidu's general-purpose agent DuMate has been upgraded to serve as the unified gateway to Baidu's full range of specialized agents, and Baidu has fully opened its Harness engine and Agent API. Over the past quarter, DuMate became the fastest-growing general-purpose agent. The DuMate team also put the three principles of organizational evolution into practice: it has iterated more than 160 times in six months, with 90% of its code and 50% of its external communication materials generated by DuMate.

KOOKO, the global AI-powered platform for documents and creative workflows, brings together publicly available information, professional databases, and users' own content to help research, write, and create high-quality artifacts such as research reports, slide decks, and websites. Upgraded from GenFlow 3.0, KOOKO draws on 1.8 billion professional resources from Baidu Wenku and Baidu Drive, 750 million Chinese and English academic papers, and a network of millions of experts.

Baidu Famou, a self-evolving agent designed to address complex operational challenges across industries, has been upgraded to version 3.0. Combining an agent platform with forward-deployed engineer (FDE) services, it takes direct responsibility for continuously optimizing final business results in real production systems. Its customers now span 25 of China's 41 industrial categories.

Cloud and Chips Strengthen the Foundation as Apollo Go Expands Globally

At the event, Baidu AI Cloud launched the Tianchi 1024-card Supernode, an all-optical interconnect supernode built for high-speed data transmission. Kunlunxin, Baidu's AI chip business, also announced a major roadmap update: the Kunlunxin M100 inference chip has shipped as planned and delivers exceptional cost-effectiveness, while the flagship M300 is expected to ship in 2027 with a major leap in computing power, memory, and interconnection. "We believe inference will be the primary driver of sustained long-term growth in the AI chip market," Li said.

"Embodied AI is now at a stage comparable to where autonomous driving stood in 2016," Li said. Apollo Go, Baidu's autonomous ride-hailing service, now covers 28 cities worldwide and has logged more than 350 million autonomous kilometers, including over 240 million in fully driverless mode. It has launched fully driverless commercial operations in Abu Dhabi and Dubai, and partnered with CAR Inc. in Hainan to pioneer the world's only driverless car rental model.

Li said intelligent organizations will open three major opportunities: asset-light operations, new competitive moats, and the creation and discovery of new markets. "So now is the moment to drive the intelligent evolution of organizations," he said.

"Baidu is ready to work with everyone to rethink, redesign, and recreate the organization," Li said. "Let us take the initiative in organizational transformation and win the future through intelligent evolution."

About Baidu

Founded in 2000, Baidu's mission is to make the complicated world simpler through technology. Baidu is a leading AI company with strong internet foundation, trading on the NASDAQ under "BIDU" and HKEX under "9888." One Baidu ADS represents eight Class A ordinary shares.

Media Contact:
intlcomm@baidu.com

Cision View original content to download multimedia:https://www.prnewswire.com/news-releases/baidu-unveils-latest-ernie-models-amid-broader-full-stack-ai-lineup-and-debuts-token-exchange-to-power-intelligent-organizations-302937959.html

SOURCE Baidu, Inc.

FAQ

AI-generated questions and answers. How Rhea-AI works. Not financial advice.

What does Baidu's ERNIE 5.5 Flash Preview API cost?

ERNIE 5.5 Flash Preview costs RMB 1 per million input tokens and RMB 4 per million output tokens through Baidu AI Cloud's Qianfan platform. Cached input costs RMB 0.04 per million tokens. The model is also available for free on OpenRouter.

What is Baidu's Token Exchange?

Token Exchange is a marketplace connecting AI model service providers with customers. It has onboarded more than 100 model resources and, one month in, handles a daily average of more than 50 million calls and over 700 billion tokens.

How does Baidu's Token Exchange check model services and providers?

Token Exchange checks resources through model fingerprint comparisons and tests of capability, performance and availability. Provider ratings use regular spot checks and real-time monitoring of calls from end to end.

Keep reading