Skip to content

Baidu / ERNIE

AI at Baidu: the ERNIE models and products, Apollo Go robotaxis and Kunlun chips.

12 picksRelated topicsQwenByteDance / DoubaoIndustry

Latest picks

1–12 of 12

Nov 12Thursday

Latent Space

Decision model roundup: Google Gemini Decision 1 handles 1,000 questions per call

Google released Gemini Decision 1 and the Gemini Decisions API, which settle closed-option decisions in a single computation instead of generating text token by token. The model handles 1,000 questions per call.

Why it matters: Google's Decision 1 leads Jev in Google's own tests, yet independent evaluations put the two level, with Decision 1 costing more per task.

Nov 7Saturday

Latent Space

Anthropic IPO outlook: revenue growth meets a valuation fight

Investors are questioning Anthropic's IPO valuation. Materials cited by the WSJ put its October annualized revenue above $80B, while some investors argue the valuation should fall from $1.5T to near $1T.

Why it matters: Anthropic's revenue is climbing while investors cut their IPO valuation expectations, a pullback that frames the pricing dispute around its listing.

Jul 24Friday

AI HOT (Curated Pool)

Baidu Dazi update: cross-device task handoff, desktop browser agent now live

Baidu Dazi now syncs task context between desktop and phone, so you can pick up where you left off across devices. The desktop version gets an embedded browser that auto-opens pages for research and downloads; the phone side supports cloud remote control. Baidu claims 20% faster task completion, 25% better token utilization, 100% simple-task accuracy, 94% complex-task delivery rate, and up to 75% lower credit cost. The post doesn't disclose testing methodology or baselines—treat the numbers as directional.

Why it matters: Baidu's Dazi hands tasks between desktop and phone, with an embedded browser that looks things up and downloads files; Baidu says average task time drops 20% and token use improves 25%.

Jun 28Sunday

Financial Times · Technology

Wuhan’s robotaxi push is pushing out ride-hailing drivers

FT reports that Wuhan’s mass deployment of robotaxis has halved ride-hailing driver incomes. Baidu’s Apollo Go runs over 1,000 driverless cars in the city, with daily peak orders exceeding 200,000 and per-km pricing undercutting Didi’s regular rides. Drivers interviewed say monthly earnings dropped from ¥8,000 to ¥3,000–4,000. The article doesn’t give a citywide unemployment figure, but it quotes multiple drivers who are quitting or selling their cars. This is FT field reporting, not an industry PR piece—just keep in mind the sample is Wuhan-specific.

Why it matters: Baidu's Apollo Go put over 1,000 robotaxis in Wuhan, peaked above 200,000 orders a day and priced below Didi Express, while local ride-hailing drivers' monthly income fell from 8,000 yuan to 3,000-4,000.

Jun 9Tuesday

Financial Times · Technology

Pentagon restores Alibaba, Baidu and BYD to Chinese military groups blacklist

The Pentagon reinstated Alibaba, Baidu, and BYD as U.S. national security risks after removing the three companies in February; the RSS snippet does not disclose the legal basis, timetable, or company responses.

Why it matters: Alibaba, Baidu and BYD are back on the Pentagon's Chinese military companies blacklist months after being removed, and the back-and-forth itself is the story.

May 18Monday

AI HOT (Curated Pool)

Baidu Core AI Business Revenue Exceeded RMB 13.6B in Q1

Baidu reported that its core AI-driven business generated more than RMB 13.6 billion in Q1 2026, up 49% year over year, and accounted for more than half of Baidu’s general business revenue for the first time.

Why it matters: Baidu's core AI-driven business passed 13.6 billion yuan in Q1 revenue, up 49% year on year and more than half of general business revenue, led by AI cloud, AI apps and Apollo Go.

Bloomberg Technology

Baidu AI Sales Eclipse Waning Legacy Ads for the First Time

Baidu reported a 1% revenue decline as growth in nascent AI businesses offset shrinking traditional internet revenue; the post does not disclose AI sales, advertising revenue, or details of the agentic AI pivot.

Why it matters: Baidu's latest earnings show AI-related revenue passing its shrinking legacy search ads, with total revenue down 1% and the mix shifting tracks.

May 13Wednesday

AI HOT (Curated Pool)

Miaoda App and Enterprise Edition launch with 90% self-generated code

Baidu launched the Miaoda app and Miaoda Enterprise Edition, saying 90% of the Miaoda app’s code was generated by Miaoda itself; Miaoda-generated apps have served over 10 million users and reached a total value of RMB 5 billion.

Why it matters: Baidu opened its Miaoda coding assistant as app and enterprise editions, saying Miaoda wrote 90% of the app's own code, with over 10 million cumulative users and 5 billion yuan in total app value.

QbitAI · WeChat

An 8-Year-Old Turns Ideas into Apps as Baidu Launches Miaoda 3.0

Baidu launched Miaoda 3.0 at its 2026 Create conference, adding iOS and Android app generation, Android packaging, online hot updates, and an enterprise edition with three-level permissions, environment isolation, and SLA commitments.

Why it matters: Baidu's Miaoda 3.0 extends no-code generation from web to iOS and Android, adds Android packaging and hot updates, and gives the enterprise edition three-tier permissions and environment isolation.

May 9Saturday

AI HOT (Curated Pool)

Baidu releases ERNIE 5.1 with compressed parameters and training cost

Baidu released ERNIE 5.1 with total parameters reduced to about one third of the original scale, active parameters to about one half, and pretraining cost to about 6% of same-scale models; the model is available on the ERNIE platform and Baidu AI Studio.

Why it matters: Baidu's ERNIE 5.1 cuts total parameters to about a third and pretraining cost to 6% of comparable models, with head-to-head numbers against DeepSeek-V4-Pro and Gemini 3.1 Pro.

AI HOT (Curated Pool)

ERNIE 5.1 Released With Pretraining Cost at 6% of Comparable Models

Baidu released ERNIE 5.1, saying it builds on ERNIE 5.0 pretraining and improves search, reasoning, knowledge QA, creative writing, and agent capabilities, with pretraining cost at about 6% of comparable models.

Why it matters: Baidu built ERNIE 5.1 on 5.0, upgrading search, reasoning, knowledge Q&A, creative writing and agent abilities, at 6% of the pretraining cost of comparable models.

May 7Thursday

Bloomberg Technology

Pentagon’s Botched Blacklist Highlights Frail Trump-Xi Truce

Bloomberg says a mishandled Pentagon blacklist of Alibaba and Baidu strained US-China tech ties. The RSS snippet does not disclose blacklist terms, timing, or company impact. The key signal is US policy conflict over China tech controls.

Why it matters: The Pentagon added Alibaba and Baidu to a list of companies tied to the military, then pulled it back, showing Washington has no single line on Chinese tech giants.