Enterprise AI in Numbers · Capability and cost
Capability of the best AI model (Epoch Capabilities Index)
168.8 points
168.8 points (Claude Opus 5.5) as of September 22, 2026, according to Epoch AI, Capabilities and benchmarking, unchanged from 168.8 points on September 3, 2026. The highest Epoch Capabilities Index (ECI) score of any model released by the date shown. The ECI combines results across many benchmarks into one scale, so models tested on different benchmarks can be compared.
What it means for leaders
The frontier keeps moving every few months, and the lead changes hands between a small group of labs. Avoid locking into one model provider; design for switching models as the best option changes.
What would change the reading: How quickly each new frontier model is matched by others.
The trend
Every reading
| Date | Value | Change | Detail | Source |
|---|---|---|---|---|
| 168.8 points | ±0.0 points | Claude Opus 5.5 | Source | |
| 168.8 points | +2.3 points | GPT-6 Astra | Source | |
| 166.5 points | +2.6 points | Claude Fable 5.1 | Source | |
| 163.9 points | +0.5 points | Claude Opus 5 | Source | |
| 163.4 points | +0.5 points | Claude Fable 5 | Source | |
| 162.9 points | +3.3 points | GPT-5.5 Pro | Source | |
| 159.6 points | +2.5 points | GPT-5.4 Pro | Source | |
| 157.1 points | +0.8 points | GPT-5.3 Codex | Source | |
| 156.3 points | +3.2 points | GPT-5.2 Pro | Source | |
| 153.1 points | +2.5 points | Gemini 3 Pro | Source | |
| 150.6 points | +0.6 points | GPT-5 Pro | Source | |
| 150.0 points | +2.7 points | GPT-5 | Source | |
| 147.3 points | +0.6 points | o3-pro | Source | |
| 146.7 points | +2.7 points | o3 | Source | |
| 144.0 points | +2.3 points | Gemini 2.5 Pro (Mar 2025) | Source | |
| 141.7 points | +6.0 points | o1 | Source | |
| 135.7 points | +5.7 points | o1-mini | Source | |
| 130.0 points | +1.0 points | Claude 3.5 Sonnet | Source | |
| 129.0 points | +1.7 points | GPT-4o (May 2024) | Source | |
| 127.3 points | +0.4 points | GPT-4 Turbo (Apr 2024) | Source | |
| 126.9 points | +0.5 points | Claude 3 Opus | Source | |
| 126.4 points | +0.7 points | GPT-4 Turbo (Nov 2023) | Source | |
| 125.7 points | +15.7 points | GPT-4 (Mar 2023) | Source | |
| 110.0 points | — | LLaMA-65B | Source |
Definition and method
- What it measures
- The highest Epoch Capabilities Index (ECI) score of any model released by the date shown. The ECI combines results across many benchmarks into one scale, so models tested on different benchmarks can be compared.
- Population
- AI models scored by Epoch AI
- Updated
- When new models are scored
- Source
- Epoch AI, Capabilities and benchmarking · CC BY 4.0
Caveats
- Epoch recalculates scores as new benchmark results arrive, so earlier values can shift slightly.
- Benchmarks measure capability on tests, not value in your workflows.
Cite this
Enterprise AI in Numbers, "Capability of the best AI model (Epoch Capabilities Index)", https://lucianoayres.com/ai-brief/data/frontier-ai-capability/, retrieved October 25, 2026. Data: Epoch AI, Capabilities and benchmarking.