Haitong International: Grok-4 Leads AI Advancement to Capitalize on Computing Power and Vertical Industry Tracks

According to Zhitong Finance APP, Haitong International released a research report stating that on July 10, Grok 4 was launched, surpassing existing models. With xAI taking the lead into the next generation of AI, it will encourage industry enterprises to actively explore the integration with cutting-edge technologies, accelerate innovation, and push the entire industry towards a higher stage of development. Cloud service providers and data center operators will directly benefit from the continuously growing demand for computing power. AI solution providers with vertical domain advantages and data barriers will stand out in the competition. Recommended targets: Dingjie Smart (300479.SZ), Foxit Software (688095.SH), Digital China (000034.SZ), iFLYTEK (002230.SZ); benefiting targets: Youke De (688158.SH), Haitai Ruisheng (688787.SH).

Haitong International's main viewpoints are as follows:

Deep thinking and group decision-making synergy, reconstructing the superhuman reasoning computing paradigm

Grok-4's reasoning computing capability has achieved a breakthrough in levels, with its pre-training computing volume and reasoning computing ability improved by more than ten times compared to previous generations, and the training scale reaching a hundred times that of Grok-2. Through testing 2,500 doctoral-level problems covering natural sciences, engineering, and other disciplines, Grok-4 achieved a score of 45% in the Human-Level Evaluation (HLE), double that of the previously most advanced AI Gemini 2.5 pro.

Grok-4 not only fully surpasses the academic capabilities of human graduate students but also refreshes records with perfect scores in authoritative benchmark tests such as GPQA and AIME25. Among them, the multi-agent collaborative Grok-4 Heavy can simultaneously combine deep thinking and group collaboration error correction capabilities, successfully achieving a perfect score in AIME25. This non-human reasoning efficiency has rendered traditional human-designed tests meaningless, and its capability boundaries are pushing forward the discovery of new technologies and physical laws, expected to generate breakthrough scientific research results within two years.

Connecting the full chain of real-world scenarios, verifying cross-industry decision execution capabilities

In terms of solving real-world scenario problems, Grok-4 has shown revolutionary progress: the voice function has doubled its response speed and halved its latency, and the Eve English synthesis technology endows conversations with natural magnetism and emotional fluidity, significantly enhancing user experience compared to competitors; in the vending machine management test (Vending-Bench), Grok-4 generated a net asset value of 4694.15, crushing the second place ClaudeOpus 4 by more than two times, verifying its long-term strategy execution capability; at the same time, after opening the 256K context API interface, it has assisted ARC Institute in screening millions of experimental data to generate research hypotheses in the biomedical field, becoming the preferred tool in financial decision-making, and even completed the autonomous development of a first-person shooter game in just 4 hours, proving its ability to integrate the toolchain throughout the process to solve complex cross-industry tasks.

Focusing on pixel-level video generation revolution, building a new ecology of human-machine collaborative perception

The downside is that Grok-4's multimodal capabilities are still a significant shortcoming. Although progress has been made in image understanding and generation, it still needs substantial improvement and has not yet achieved human-level audiovisual perception and interaction capabilities. **The next generation of research and development will focus on breakthroughs in video generation technology through " The end-to-end training of "input pixels - output pixels" has achieved a closed loop in AI video creation on platform X, with plans to launch an integrated 3D resource automatic generation system using Unreal Engine next year to empower the gaming and film industries. ** In the short term, the focus will be on strengthening dedicated programming models and optimizing image recognition technology, with the ultimate goal of building a super intelligent agent that combines deep thinking, real-time response, and multimodal collaboration, fundamentally reshaping the paradigm of human-machine collaboration.

Risk Warning

Intensifying technological competition, insufficient computing power supply, and data privacy compliance risks