Charts / New Enterprise Associates
Open models are closing the gap at a fraction of the compute
Scatter plot comparing MMLU accuracy against training compute (FLOP) for open and closed AI models, demonstrating that open models like DeepSeek-V2 and Gemma 2 9B achieve high accuracy at 6x to 7x less compute than earlier models.
| Label | Value shown on the chart |
|---|---|
| DeepSeek-V2 | 7x less compute |
| Gemma 2 9B | 6x less compute |
Shown as a preview with credit and a link to the original. Publishers can request removal.
