LLM Leaderboard
Analyze and compare API models across benchmarks, pricing, and capabilities.
LLM Rankings
Best models and API providers in each category
One API. Every model.
Test in our playground. Deploy with our unified API. Access 100+ models through a single OpenAI-compatible endpoint.
Context Window
Maximum input context length for each model
While tokenization varies between models, on average, 1 token ≈ 3.5 characters in English.Note: Each model uses its own tokenizer, so actual token counts may vary significantly.
As a rough guide, 1 million tokens is approximately equivalent to:
30 hours
of a podcast
~150 words per minute
1,000 pages
of a book
~500 words per page
60,000 lines1
of code
~60 characters per line
[1] Based on average characters per line. See Wikipedia.
Comparisons
Capabilities vs price across models (GPQA vs $/1M input tokens); plus other comparisons
API Providers - Open LLM Providers
Price and performance across providers for Llama 4 Maverick
Provider performance varies significantly. Some providers run full-precision models on specialized hardware accelerators (like Groq's LPU or Cerebras' CS-3), while others may use quantization (4-bit, 8-bit) to simulate faster speeds on commodity hardware. Check provider documentation for specific hardware and quantization details, as this can impact both speed and model quality.
Trends
Tracking AI progress across nations, model types, and organizations
AI Progress: US vs China
SOTA comparison over time between the United States and China.
Open vs Closed Models
Comparing SOTA progression of open-source and proprietary models over time.
Organization Progress
SOTA progression by organization over time.
Observe how different processing speeds affect real-time token generation.Try adjusting the speeds using the number inputs above each panel ↑
Values reset every 5 seconds to demonstrate different speeds
Popular Model Comparisons
Gemini 2.5 Pro
Claude Sonnet 4