Meituan LongCat Team Launches General 365: A New Benchmark Revealing the Limits of AI Reasoning Capabilities
The Meituan LongCat team has officially released "General 365," a rigorous new benchmark designed to evaluate the reasoning capabilities of large language models. In an extensive assessment involving 26 mainstream AI models, the results highlight a significant performance gap in the industry. Gemini 3 Pro, identified as the top-performing model in this evaluation, achieved an accuracy rate of only 62.8%. Notably, the vast majority of the models tested failed to reach the 60% "passing line," suggesting that complex reasoning remains a formidable challenge for current artificial intelligence. This benchmark establishes a new standard for measuring the logical depth and accuracy of next-generation AI systems, providing a clear look at the current ceiling of model performance.
Key Takeaways
- New Reasoning Standard: Meituan's LongCat team has open-sourced General 365, a benchmark specifically focused on evaluating AI reasoning.
- Extensive Model Testing: The initial evaluation included 26 of the world's most prominent mainstream AI models.
- Gemini 3 Pro Leads: Currently the highest-scoring model, Gemini 3 Pro achieved an accuracy of 62.8% on the benchmark.
- Widespread Performance Gap: The majority of tested models were unable to reach a 60% accuracy threshold, indicating significant room for improvement in AI reasoning.
In-Depth Analysis
The General 365 Benchmark: A New Yardstick for AI
The release of General 365 by the Meituan LongCat team represents a significant shift in how large language models (LLMs) are evaluated. By focusing specifically on reasoning, this benchmark moves beyond general conversational fluency to test the logical core of AI systems. The evaluation of 26 different mainstream models provides a comprehensive cross-section of the current AI landscape. The fact that such a wide variety of models were tested suggests that General 365 is intended to be a universal standard, applicable across different architectures and development philosophies. The results of this broad test serve as a reality check for the industry, grounding the excitement surrounding AI in hard, comparative data.
Analyzing the Performance Ceiling: Gemini 3 Pro
According to the data provided by the Meituan LongCat team, Gemini 3 Pro currently stands as the "strongest on Earth" within the context of this benchmark. However, its accuracy score of 62.8% is revealing. While it holds the top position among the 26 models tested, a score in the low 60s suggests that even the most advanced models are struggling with the specific reasoning challenges posed by General 365. This 62.8% figure sets a clear performance ceiling for the current generation of AI, indicating that while the technology is leading the field, it has yet to achieve mastery over complex reasoning tasks. This specific data point provides a crucial metric for developers aiming to surpass the current state-of-the-art capabilities.
The 60% Threshold: A Challenge for the Majority
One of the most striking findings from the General 365 evaluation is the failure of the vast majority of models to reach the 60% accuracy mark. In many academic and professional contexts, 60% is considered the minimum "passing" grade. The fact that most mainstream models failed to "touch the passing line" highlights a systemic weakness in current AI reasoning. This widespread inability to meet a basic threshold of accuracy on General 365 suggests that the benchmark is designed with a high degree of difficulty, or that current model training methodologies are not yet sufficiently optimized for the types of reasoning tasks included in this set. This gap between the top performers and the rest of the field underscores the difficulty of scaling reasoning capabilities across different AI platforms.
Industry Impact
The introduction of General 365 by Meituan is likely to have a lasting impact on the AI industry by redefining the criteria for "high-performance" models. By establishing a benchmark where even the top model scores only 62.8%, Meituan has created a "high-bar" environment that encourages more rigorous development. This will likely shift the industry's focus toward improving the logical consistency and deep reasoning of models rather than just increasing their parameter size or conversational variety. Furthermore, as an open-source tool, General 365 provides a transparent and standardized way for developers to measure their progress against the current industry leaders like Gemini 3 Pro, potentially accelerating the pace of innovation in AI reasoning.
Frequently Asked Questions
Question: What is the highest score achieved on the General 365 benchmark so far?
As of the latest report from the Meituan LongCat team, the highest accuracy score achieved is 62.8%, which was reached by Gemini 3 Pro.
Question: How many models were tested in the initial General 365 evaluation?
The Meituan LongCat team tested a total of 26 mainstream AI models to establish the initial performance data for the General 365 benchmark.
Question: What was the general performance level of the models tested?
The majority of the 26 models tested were unable to reach the 60% accuracy threshold, which is described as the "passing line" for the benchmark.
