Meituan LongCat Releases General 365: A New Benchmark Revealing Significant Gaps in AI Reasoning Capabilities
The Meituan LongCat team has officially launched General 365, a rigorous new benchmark designed to evaluate the reasoning capabilities of Large Language Models (LLMs). In an initial assessment of 26 mainstream models, the results underscore a significant performance gap in the industry. Gemini 3 Pro, currently recognized as one of the most powerful models, led the group but only achieved an accuracy rate of 62.8%. Most notably, the vast majority of the 26 models tested failed to reach the 60% passing threshold. This benchmark aims to establish a more demanding standard for AI evaluation, highlighting that complex logical reasoning remains a major challenge even for state-of-the-art systems.
Key Takeaways
- New Reasoning Standard: Meituan's LongCat team has introduced General 365, a benchmark specifically focused on testing the reasoning depth of AI models.
- Gemini 3 Pro Performance: As the top performer in the initial test, Gemini 3 Pro reached an accuracy of 62.8%, setting the current ceiling for the benchmark.
- Widespread Failure to Pass: The majority of the 26 mainstream models evaluated were unable to meet the 60% accuracy mark, which serves as the benchmark's passing line.
- Industry Benchmark Shift: General 365 is positioned as a 'new yardstick' for reasoning, suggesting that current evaluation methods may not be sufficiently challenging for modern LLMs.
In-Depth Analysis
The Launch of General 365 and the 26-Model Evaluation
The Meituan LongCat team has officially entered the AI evaluation space with the release of General 365. This benchmark is not merely another test suite but is described as a 'new yardstick' for reasoning evaluation. The team conducted a comprehensive assessment involving 26 of the most prominent mainstream models currently available in the industry. This broad scope of testing provides a clear snapshot of where the AI field stands regarding logical processing and complex reasoning tasks.
The methodology behind General 365 appears to be designed with a high degree of difficulty. By testing a wide array of 26 models, the LongCat team has ensured that the results are representative of the current technological landscape. The focus on 'reasoning' specifically targets one of the most critical and difficult aspects of artificial intelligence development, moving beyond simple language generation or information retrieval into the realm of cognitive logic.
Analyzing the Performance Gap: Gemini 3 Pro and the 60% Threshold
The results released by the LongCat team reveal a stark reality for the AI industry. Gemini 3 Pro, which is identified as the 'strongest model on the planet' within the context of this evaluation, achieved an accuracy rate of 62.8%. While this score places it at the top of the 26 models tested, it also highlights the extreme difficulty of the General 365 benchmark. An accuracy rate of just over 60% for a leading model suggests that the tasks within General 365 are significantly more complex than those found in traditional benchmarks.
Perhaps more concerning for the industry is the performance of the remaining models. The LongCat team reported that the vast majority of the 26 models failed to even reach the 60% 'passing' mark. This widespread failure to hit a basic competency threshold indicates that reasoning remains a significant bottleneck for most Large Language Models. The data suggests that while AI has made strides in many areas, the ability to consistently and accurately reason through complex problems is still lacking in most mainstream architectures. This gap between the 'strongest' model and the rest of the field, combined with the low overall scores, defines the current frontier of AI research.
Industry Impact
The introduction of General 365 by Meituan's LongCat team is set to have a profound impact on how AI progress is measured. By establishing a benchmark where even the most advanced models struggle to exceed a 60% accuracy rate, Meituan is pushing the industry toward more rigorous development standards. This 'new yardstick' serves as a critical reminder that current AI capabilities, while impressive, still fall short of reliable logical reasoning.
For the AI industry, the General 365 results provide a clear roadmap for future improvement. The fact that 26 mainstream models were tested and most failed to pass the 60% mark will likely spur a shift in focus from model size and parameter count to the quality of reasoning and logical consistency. As developers strive to climb the rankings of General 365, the benchmark will likely become a key metric for determining the true 'intelligence' of future Large Language Models, moving the goalposts closer to achieving genuine Artificial General Intelligence (AGI).
Frequently Asked Questions
Question: What is the primary purpose of the General 365 benchmark?
General 365 was released by the Meituan LongCat team to serve as a new, more rigorous standard for evaluating the reasoning capabilities of AI models. It aims to provide a realistic measure of how well models can handle complex logical tasks.
Question: How did the top models perform on this new test?
Gemini 3 Pro was the highest-performing model among the 26 tested, achieving an accuracy of 62.8%. However, the majority of the other mainstream models failed to reach the 60% passing threshold, indicating that the benchmark is highly challenging.
Question: Who developed General 365 and how many models were involved in the initial report?
General 365 was developed and released by the Meituan LongCat team. Their initial report included performance data for 26 different mainstream AI models.
