Meituan LongCat Team Launches General 365: A Rigorous New Benchmark for AI Reasoning Evaluation

Key Takeaways

• Meituan's LongCat team has introduced General 365, a new benchmark specifically for reasoning evaluation.
• Evaluation of 26 mainstream AI models shows a significant struggle with the benchmark's requirements.
• Gemini 3 Pro emerged as the leader but only secured a 62.8% accuracy rate.
• Most tested models failed to reach the 60% passing threshold, indicating a high level of difficulty.

In-Depth Analysis

The Introduction of General 365

The Meituan LongCat team has officially unveiled General 365, a benchmark that seeks to redefine how reasoning is measured in the field of artificial intelligence. By positioning this tool as a "new scale" for reasoning evaluation, the team addresses the need for more rigorous testing environments. The release comes at a time when many models claim high performance on existing benchmarks, yet General 365 suggests that true reasoning capabilities may still be lacking across the industry. The focus of this benchmark is to provide a clear, quantifiable measure of how well models can navigate complex logical tasks.

Benchmarking the Giants: Gemini 3 Pro and the 60% Threshold

The results from the initial testing phase of General 365 provide a sobering look at the current state of AI. Out of 26 mainstream models evaluated, the performance levels were notably lower than what is often seen in other standardized tests. Gemini 3 Pro, which is currently regarded as one of the most powerful models available globally, managed to achieve an accuracy of 62.8%. While this placed it at the top of the list, the score itself highlights the difficulty of the General 365 benchmark.

Perhaps more significant is the finding that the vast majority of the 26 models could not reach the 60% mark. In the context of this evaluation, 60% is treated as the "passing line," and the failure of most mainstream models to meet this standard suggests that current reasoning architectures are not yet optimized for the specific challenges presented by General 365. This gap between the leading model and the rest of the field, combined with the overall low scores, underscores the rigorous nature of Meituan's new evaluation framework.

Industry Impact

The release of General 365 by Meituan's LongCat team is significant for the AI industry as it introduces a more stringent set of criteria for reasoning. By demonstrating that even the most advanced models like Gemini 3 Pro struggle to exceed a 60% accuracy rate by a wide margin, the benchmark exposes the limitations of current large language models. This is likely to push developers and researchers to focus more heavily on the logical and reasoning components of AI, rather than just linguistic fluency or pattern recognition. As a new "ruler" for the industry, General 365 provides a clear target for future model iterations and sets a high bar for what constitutes a "passing" grade in AI reasoning.

Frequently Asked Questions

Question: What is the primary purpose of Meituan's General 365?

General 365 is a reasoning evaluation benchmark released by the Meituan LongCat team. Its purpose is to serve as a new, more rigorous standard for measuring the reasoning capabilities of mainstream AI models.

Question: How did the top-performing models fare on this benchmark?

In a test of 26 mainstream models, Gemini 3 Pro performed the best with an accuracy rate of 62.8%. However, most other models failed to reach the 60% passing threshold, indicating that the benchmark is highly challenging for current AI technology.

Question: How many models were included in the initial General 365 evaluation?

The Meituan LongCat team tested a total of 26 mainstream models to establish the initial performance data for the General 365 benchmark.