Lead: This November 2 briefing has seven items: an agent benchmark pours cold water on the industry, research agents and a finance model both make progress on the same day, embodied AI lands two fundings, and an AI forum in Guangdong unveils several results at once.
1. Agent benchmark GDPx launches; the top score is only 12%
SemiAnalysis released GDPx on November 2, covering 11 domains and 206 tasks. Claude Opus 5.5 leads at 12.0%, followed by Gemini 4 Carbon at 10.9% and GPT-6 Astra at 9.6%; finance and legal score zero under binary grading.
2. Tencent Hunyuan releases Hyra 2.0, letting agents build research tools
Tencent Hunyuan released Hyra 2.0 on November 2, shifting the focus from solving problems directly to letting agents develop reusable scientific models and tools. It has already produced HyInv and the quantum error-correction tool HyQEC; HyInv has only 1.96 million parameters, generates 57 times faster than ProteinMPNN, and the results are public on GitHub.
3. Xuanyuan 4.5 launches, improving finance scenarios by 30%
Du Xiaoman, the Hong Kong University of Science and Technology, and the Hong Kong Generative AI Research and Development Centre released Xuanyuan 4.5 at Hong Kong FinTech Week on November 2, calling it the first self-evolving finance-enhanced large model. It improves results by 30% in scenarios such as customer acquisition, risk control, and customer operations, at about 10% of the cost of general-purpose models.
4. RoboScience closes a Series A2 round worth hundreds of millions of yuan
RoboScience announced on November 2 that it had closed a Series A2 round worth hundreds of millions of yuan, co-led by Amber Capital, Hongruida Investment, and Shanghai Semiconductor Industry Investment. The funding will be used to iterate its general embodied model Visics and to advance validation and scaled deployment in logistics, retail, and industrial scenarios.
5. GenRobot discloses a $100 million Series A+ round
GenRobot disclosed on November 2 that it had closed a $100 million Series A+ round in the third quarter of this year, with investors including industrial investors such as SAIC Group and Baidu Group as well as several financial institutions. The company turns real human operations into robot training data and has worked with more than 30 AI companies.
6. Greater Bay Area AI forum unveils several results at once
On the morning of November 2, the AI sub-forum of the Greater Bay Area Science Forum was held in Nansha, Guangzhou. Guangzhou National Laboratory released EpiETL, an epidemic intelligence platform; Pengcheng Laboratory and Beijing University of Posts and Telecommunications released a 6G semantic communication base station prototype; and XPeng released a full-stack solution for the physical world.
7. Report says OpenAI may rename data centers “AI compute campuses”
According to Axios, OpenAI is preparing a public-relations push after the US midterm elections, considering renaming data centers “AI compute campuses” and promoting ideas such as AI “speed limits.” TechCrunch relayed the report on November 1, saying data centers are unpopular locally.
