Historical Token Throughput per GPU
Historical Token Throughput per GPU
- H100 FP8
- H100 FP16
- MI300X FP8
AI data
Capability
Model, system, throughput, efficiency, context, and agentic benchmark progress.
Performance
Historical Inference Speeds
Historical Inference Speeds
- Groq
- OpenAI
- Anthropic
Models
Historical Token Efficiency Across Frontier Models
Historical Token Efficiency Across Frontier Models
- GPT-5.5
- Claude 4.5
- Gemini 3
Epoch Capabilities Index (ECI)
Epoch Capabilities Index (ECI)
- Composite
- Coding
- Math
Frontier Model Training Compute
Frontier Model Training Compute
- Frontier maximum
- Frontier median
Frontier Model Context Window Growth
Frontier Model Context Window Growth
- Maximum shipped
- Frontier median
Agentic and Tool-Use Benchmarks
Agentic and Tool-Use Benchmarks
- SWE-bench Verified
- OSWorld
- GAIA