Updated May 2026
LLM Leaderboard 2026
Independent ranking of GPT, Claude, Gemini, DeepSeek, Kimi and 300+ models — composite score updated from llm-stats.com benchmarks and live API metrics.
Current leaders
Full Leaderboard
Top 20 models| # | Model | LLM Stats | Reasoning | Coding | Agent | Code Arena | Context | Speed | Pricing $/M | License |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Mythos PreviewPreview Anthropic | 70.1 | 71.3 | 57.3 | 48.8 | — | — | — | $36.11 | Prop. |
| 2 | GPT-5.5New OpenAI | 64.0 | 62.9 | 53.1 | 44.0 | 1,267 | 1.1M | 124c/s | $7.78 | Prop. |
| 3 | Claude Opus 4.7 Anthropic | 61.1 | 62.7 | 51.6 | 42.3 | 1,837 | 1.0M | 63c/s | $7.22 | Prop. |
| 4 | GPT-5.4 OpenAI | 61.0 | 57.9 | 44.3 | 37.8 | 1,717 | 1.0M | 224c/s | $3.89 | Prop. |
| 5 | GPT-5.2 Pro OpenAI | 61.0 | 56.5 | — | 29.9 | — | 400K | — | $37.33 | Prop. |
| 6 | Kimi K2.6New Moonshot AI | 58.7 | 59.5 | 45.6 | 38.7 | 1,224 | 262K | 40c/s | $1.29 | Open |
| 7 | Gemini 3.1 Pro Google | 57.8 | 59.0 | 44.1 | 33.8 | 2,093 | 1.0M | — | $3.89 | Prop. |
| 8 | Claude Opus 4.6 Anthropic | 57.5 | 59.9 | 45.6 | 38.1 | 2,003 | 1.0M | 99c/s | $7.22 | Prop. |
| 9 | Seed 2.0 Pro ByteDance | 56.6 | 54.5 | 33.3 | 29.3 | — | — | — | — | Prop. |
| 10 | GPT-5.2 OpenAI | 56.1 | 54.2 | 35.7 | 26.4 | 1,520 | 400K | 173c/s | $3.11 | Prop. |
| 11 | Gemini 3 Pro Google | 56.0 | 50.2 | 33.4 | 24.1 | 1,579 | — | — | — | Prop. |
| 12 | Gemini 3 Flash Google | 54.7 | 49.7 | 31.5 | 25.6 | 1,689 | 1.0M | 300c/s | $0.78 | Prop. |
| 13 | GPT-5.1 Instant OpenAI | 54.6 | 49.6 | 31.3 | — | 777 | 400K | 209c/s | $2.22 | Prop. |
| 14 | GPT-5.1 OpenAI | 53.3 | 48.0 | 31.8 | — | 1,225 | 400K | 447c/s | $2.22 | Prop. |
| 15 | GPT-5 Medium OpenAI | 53.1 | 44.3 | — | — | 1,095 | 400K | 100c/s | $2.22 | Prop. |
| 16 | Muse Spark Meta | 53.0 | 53.0 | 32.9 | 25.0 | — | — | — | — | Prop. |
| 17 | GPT-5.1 High OpenAI | 52.0 | 52.9 | — | — | 1,140 | — | — | — | Prop. |
| 18 | DeepSeek-V4-Pro-MaxNew DeepSeek | 51.9 | 57.7 | 45.0 | 36.8 | 505 | 1.0M | 53c/s | $1.93 | Open |
| 19 | GPT-5.1 Thinking OpenAI | 51.8 | 46.2 | 30.8 | — | 1,004 | 400K | 156c/s | $2.22 | Prop. |
| 20 | Qwen3.6 Plus Alibaba Cloud | 51.1 | 52.4 | 43.3 | 31.9 | — | 1.0M | — | $0.78 | Prop. |
Source: llm-stats.com. Data refreshed on every page load.
Benchmark Glossary
LLM Stats Score
Composite score blending GPQA Diamond, SWE-Bench Verified, coding-arena ELO, output throughput, and per-token pricing into a single comparable number.
GPQA Diamond
Graduate-level science questions curated by domain experts. Tests advanced multi-step reasoning across physics, chemistry, and biology.
SWE-Bench Verified
Real GitHub issues from popular Python repos that the model must resolve end-to-end. Measures agentic software-engineering ability.
Coding Arena
Head-to-head ELO rating from pairwise coding task evaluations. Higher score means consistently preferred code output vs. competing models.
Context Window
Maximum tokens the model can process in a single request, combining input, tool calls, and output. Larger windows support longer documents and conversations.
Speed (tok/s)
Output throughput measured as tokens per second via the provider's public API, averaged over a 7-day rolling window. Crucial for streaming UIs and agentic loops.