Enterprise AI in Numbers · Capability and cost
Labs statistically tied at the AI frontier
3
3 (Anthropic, OpenAI, Google DeepMind) as of September 30, 2026, according to Epoch AI, Epoch Capabilities Index, up from 2 on September 22, 2026. The number of organizations whose best model has an ECI confidence interval that reaches the top model's score, so the data can't rank them apart.
What it means for leaders
When several labs are tied at the top, buyers have leverage: switching costs, not capability, decide the vendor. Keep contracts and integrations portable across at least two of them.
What would change the reading: A count of one (a single lab pulling ahead) or a lab from outside the US joining the tie.
The trend
Every reading
| Date | Value | Change | Detail | Source |
|---|---|---|---|---|
| 3 | +50.0% | Anthropic, OpenAI, Google DeepMind | Source | |
| 2 | ±0.0% | Anthropic, OpenAI | Source | |
| 2 | ±0.0% | OpenAI, Anthropic | Source | |
| 2 | +100.0% | Anthropic, OpenAI | Source | |
| 1 | −66.7% | OpenAI | Source | |
| 3 | +50.0% | OpenAI, Anthropic, Google DeepMind | Source | |
| 2 | +100.0% | OpenAI, Anthropic | Source | |
| 1 | −50.0% | OpenAI | Source | |
| 2 | +100.0% | Google DeepMind, OpenAI | Source | |
| 1 | −50.0% | OpenAI | Source | |
| 2 | +100.0% | OpenAI, xAI | Source | |
| 1 | −50.0% | OpenAI | Source | |
| 2 | +100.0% | OpenAI, Google DeepMind | Source | |
| 1 | −66.7% | Google DeepMind | Source | |
| 3 | +50.0% | OpenAI, Anthropic, Alibaba | Source | |
| 2 | +100.0% | OpenAI, Anthropic | Source | |
| 1 | −66.7% | OpenAI | Source | |
| 3 | +50.0% | OpenAI, Google DeepMind,Google, Anthropic | Source | |
| 2 | +100.0% | OpenAI, Anthropic | Source | |
| 1 | −66.7% | OpenAI | Source | |
| 3 | +50.0% | Anthropic, OpenAI, Meta AI | Source | |
| 2 | −33.3% | Anthropic, OpenAI | Source | |
| 3 | +50.0% | OpenAI, Google DeepMind, Anthropic | Source | |
| 2 | −50.0% | OpenAI, Anthropic | Source | |
| 4 | +33.3% | OpenAI, Anthropic, DeepSeek, | Source | |
| 3 | ±0.0% | OpenAI, Anthropic, | Source | |
| 3 | +50.0% | Anthropic, OpenAI, | Source | |
| 2 | +100.0% | OpenAI, | Source | |
| 1 | ±0.0% | OpenAI | Source | |
| 1 | — | Meta AI | Source |
Definition and method
- What it measures
- The number of organizations whose best model has an ECI confidence interval that reaches the top model's score, so the data can't rank them apart.
- Population
- Models scored on the Epoch Capabilities Index (ECI)
- Updated
- When new models are scored
- Source
- Epoch AI, Epoch Capabilities Index · CC BY 4.0
Caveats
- Ties follow Epoch AI's 90% confidence intervals; a lab just outside them may be close in practice.
- Only labs whose models Epoch has scored can count.
Cite this
Enterprise AI in Numbers, "Labs statistically tied at the AI frontier", https://lucianoayres.com/ai-brief/data/frontier-contenders/, retrieved October 25, 2026. Data: Epoch AI, Epoch Capabilities Index.