The world's leading AI models, ranked
Aggregated live from Artificial Analysis and OpenRouter. Track every frontier model: intelligence, coding, math, speed and cost, all in one place.
| Rank | Model | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5 (Adaptive Reasoning, Max Effort) Anthropic | 68.4 | 63.1 | 78 | 49 | $10 | ||||
| 2 | Claude Opus 5 (Adaptive Reasoning, Xhigh Effort) Anthropic | 67.6 | 62.5 | 77 | 48 | $10 | ||||
| 3 | Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) Anthropic | 67.2 | 62.1 | 76.5 | 65 | $20 | ||||
| 4 | Claude Opus 5 (Adaptive Reasoning, High Effort) Anthropic | 66.8 | 61.5 | 76.5 | 48 | $10 | ||||
| 5 | GPT-5.6 Sol (max) OpenAI | 66.7 | 60.9 | 77.4 | 65 | $11.25 | ||||
| 6 | Grok 4.6 (high) SpaceXAI | 66.5 | 60.9 | 76.8 | 62 | $3 | ||||
| 7 | GPT-5.6 Sol (xhigh) OpenAI | 65.8 | 59 | 78.3 | 63 | $11.25 | ||||
| 8 | Kimi K3 (max) Kimi | 65.5 | 59.7 | 76.2 | 40 | $6 | ||||
| 9 | GPT-5.6 Sol OpenAI📄🖼T | 64.3 | 57.3 | 77.2 | 64 | $11.25 | 1.1M | |||
| 10 | GPT-5.6 Terra (max) OpenAI | 63.7 | 56.6 | 76.7 | 109 | $4.5 |
Frequently asked questions
＋ What is the best AI model right now?
By the DataCore composite score (intelligence, coding, math), Claude Opus 5 (Adaptive Reasoning, Max Effort) currently leads our ranking of 584 AI models from 65 providers, updated continuously.
＋ Which AI model is the cheapest?
Grok 4.6 (high) is among the lowest-cost frontier options, around $3 per 1M output tokens. You can sort and filter the leaderboard by price.
＋ What is the fastest LLM?
Celeris-1 reaches the highest output speed (~1729 tokens/sec). Speed and latency are measured by Artificial Analysis.
＋ What is an AI model "time horizon"?
It's METR's metric: the length of task (in human time) a model can complete with a 50% success rate. A longer time horizon means the model can handle longer, more complex agentic tasks.
＋ What is the best open-source LLM?
The board tracks 231 open-weight models such as DeepSeek, Qwen, Llama and GLM. Use the "Open weights" filter to compare them head-to-head.
＋ Where does the data come from and how often is it updated?
Aggregated from Artificial Analysis (intelligence, coding, math, speed, price), METR (time horizon) and OpenRouter (catalog, live pricing). Data refreshes continuously (≈10-minute cache).
Data sources
- OpenRoutercached 20h ago
Model catalog, context window, modalities, live pricing
- Artificial Analysiscached 20h ago
Intelligence Index, coding & math scores, speed, latency
50% task-completion time horizon (how long a task a model can finish)
- Aider Polyglotcached 20h ago
Polyglot code-editing benchmark (pass rate %)