The AI model landscape continues to evolve rapidly. Here's the current state of the LLM leaderboard and what it means for developers and users.
Top 10 LLMs - March 2026
| Rank | Model | Provider | Elo Score | Benchmark | Context | Pricing (per 1M) | License |
|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.6 | Anthropic | 2,002 | 91.3% coding | 200K | $5.00 / $25.00 | Proprietary |
| 2 | Gemini 3.1 Pro | 1,855 | 94.3% coding | 1.0M | $2.50 / $15.00 | Proprietary | |
| 3 | GPT-5.4 | OpenAI | 1,665 | 92.8% coding | 1.0M | $2.50 / $15.00 | Proprietary |
| 4 | GLM-5 | Zhipu AI | 1,600 | — / 77.8% | 200K | $1.00 / $3.20 | Open Source |
| 5 | Claude Opus 4.5 | Anthropic | 1,582 | 87.0% coding | 200K | $5.00 / $25.00 | Proprietary |
| 6 | Gemini 3 Flash | 1,581 | 90.4% coding | 1.0M | $0.50 / $3.00 | Proprietary | |
| 7 | Gemini 3 Pro | 1,579 | 91.9% coding | — | — | Proprietary | |
| 8 | GPT-5.2 | OpenAI | 1,505 | 92.4% coding | 400K | $1.75 / $14.00 | Proprietary |
| 9 | Kimi K2.5 | Moonshot AI | 1,448 | 87.6% coding | 262K | $0.60 / $2.50 | Open Source |
| 10 | Claude Sonnet 4.6 | Anthropic | 1,371 | 89.9% coding | 200K | $3.00 / $15.00 | Proprietary |
Key Takeaways
1. Anthropic Takes the Lead
Claude Opus 4.6 has claimed the top spot with an impressive 2,002 Elo score and 91.3% on coding benchmarks. This marks a significant shift from late 2025 when GPT models dominated the rankings.
2. Google's Gemini 3 Family is Competitive
Gemini 3.1 Pro sits at #2 with the highest coding benchmark score (94.3%) and a massive 1M context window. Gemini 3 Flash offers exceptional value at just $0.50/$3.00 per million tokens.
3. Open Source is Closing the Gap
GLM-5 from Zhipu AI cracks the top 5 with open-source licensing. At $1.00/$3.20 per million tokens, it's competitive with proprietary models while offering the flexibility of open weights.
Qwen3.5-397B-A17B (Alibaba) sits at #12 with strong performance and open-source availability.
4. Value Leaders
| Use Case | Best Value | Why |
|---|---|---|
| Cost-conscious coding | Gemini 3 Flash | $0.50 input, 90.4% coding |
| Long context | Gemini 3.1 Pro | 1M context, top-tier performance |
| Best overall | Claude Opus 4.6 | #1 ranking, excellent coding |
| Open source | GLM-5 | Top 5, open weights, low cost |
| Balanced | Claude Sonnet 4.6 | Fast, capable, reasonable price |
Notable Movers
Rising
- GLM-5 - Chinese open-source model breaks into top 5
- Kimi K2.5 - Moonshot AI's model at #9 with aggressive pricing
- Gemini 3 Flash - Best price-to-performance ratio for production
Falling
- GPT-5.1/5.2 - Overtaken by Claude and Gemini in raw benchmarks
- Older Claude models - Opus 4.0 and Sonnet 4.0 pushed down rankings
What This Means for Q4KM Users
When choosing a model through Q4KM.ai, consider:
- For coding tasks: Claude Opus 4.6 or Gemini 3.1 Pro
- For long documents: Gemini 3.1 Pro (1M context)
- For cost-sensitive apps: Gemini 3 Flash or Kimi K2.5
- For self-hosting: GLM-5 or Qwen3.5-397B
Looking Ahead
The gap between proprietary and open-source models continues to narrow. With Chinese models like GLM-5 and Kimi K2.5 performing well, expect increased competition and potentially lower prices across the board.
Rankings based on llm-stats.com aggregated benchmarks including coding, math, reasoning, and chat evaluations. Prices and rankings may change. Last verified: March 24, 2026.