Top 20Updated 2026年3月18日
March 2026 Comprehensive AI Model Ranking
Current comprehensive LLM ranking based on llm-stats.com (as of March 2026), mainly referencing Code Arena and various benchmark scores.
Current #1
Claude Opus 4.6
Interesting Facts & Summary
This month's ranking shows fierce competition in the top tier between Anthropic and Google, alternating leads in coding capabilities and overall metrics. Meanwhile, Chinese domestic models (Zhipu, Moonshot, Alibaba) show strong performance, securing spots in the global top 20 and establishing moats in Chinese contexts and cost-efficiency.
LLMs are rapidly evolving from "probability generation" to "deep logical reasoning", marking an unprecedented pace of AI evolution.
| Rank | Model Name | Vendor/Institution | Core Advantage Metrics |
|---|---|---|---|
Claude Opus 4.6 | Anthropic | Code Arena: 2,002 / GPQA: 91.3% | |
Gemini 3.1 Pro | SWE-bench Verified: 80.6% | ||
GPT-5.4 | OpenAI | Throughput (286c/s) & Reasoning Balance | |
| 4 | GLM-5 | 智谱 AI (Zhipu AI) | One of China's strongest open-source models |
| 5 | Claude Opus 4.5 | Anthropic | High Math Reasoning (AIME 2025: 80.9%) |
| 6 | Gemini 3 Pro | GPQA Perfect Score (100%) | |
| 7 | Gemini 3 Flash | Ultra-fast Response (342c/s) & Multimodal Balance | |
| 8 | GPT-5.2 | OpenAI | AIME 2025 Reasoning Capability (80.0%) |
| 9 | Kimi K2.5 | 月之暗面 (Moonshot AI) | Long-context & Comprehensive Chinese (GPQA: 87.6%) |
| 10 | Claude Sonnet 4.6 | Anthropic | Software Engineering (SWE-bench: 58.3%) |
| 11 | GPT-5 High | OpenAI | Early GPT-5 High-performance Version |
| 12 | Qwen3.5-397B-A17B | 阿里巴巴 (Qwen) | Domestic Leader in Ultra-large Open-source Parameters |
| 13 | GLM-4.6 | 智谱 AI (Zhipu AI) | Stable Multitasking Processing |
| 14 | GPT-5.2 Codex | OpenAI | Specialized Version Optimized for Coding Tasks |
| 15 | GPT-5.1 | OpenAI | Balanced Logical Processing |
| 16 | GPT-5.1 High | OpenAI | Enhanced Logic Version |
| 17 | Qwen3.5-122B-A10B | 阿里巴巴 (Qwen) | Mid-size but Highly Efficient Open-source Model |
| 18 | Claude Sonnet 4.5 | Anthropic | Excellent Instruction Following |
| 19 | GPT-5 Medium | OpenAI | Cost-performance Balanced Version |
| 20 | GPT-5.3 Codex | OpenAI | Latest Coding Specialized Version |
List Stats
Views15
Items20
CategoryTech & Business
Comments
No comments yet. Be the first to comment!
Related Rankings
Tech & Business52 views
Global Richest People Top20
Tech & Business34 views
Top 10 Countries by Renewable Energy Installed Capacity
Tech & Business29 views
Top 12 Companies by Annual Patent Licensing Revenue
Tech & Business26 views
Top 10 Global Unicorns by Valuation
Tech & Business25 views
Top 10 Global Cities for Biotech Venture Capital Investment
Tech & Business25 views