Notchi
Download
| Score | Calculation | Displayed result |
|---|---|---|
| Speed | Uses median_time_to_first_answer_token: seconds from sending the request to receiving the first answer token. Lower is better. Values are sorted from fastest to slowest and divided at the 20th, 40th, 60th, and 80th percentiles. |
1–5. The fastest band is 5; the slowest is 1. |
| Intelligence | Uses the Artificial Analysis Intelligence Index. Higher is better. Values are sorted from highest to lowest and divided at the 20th, 40th, 60th, and 80th percentiles. | 1–5. The highest band is 5; the lowest is 1. |
| Cost efficiency | ratio = Intelligence Index ÷ blended price per 1M tokensscore = round(100 × number of reference ratios below this ratio ÷ total reference ratios)The blended price assumes three input tokens for every one output token. |
0–100. Free models score 100. If Intelligence or price is missing, no score is shown. |
Comparison range
Scores are calculated against the models in Notchi’s current model lineup. If fewer than eight models have the required data for a measure, all measured models in the Artificial Analysis catalog are used instead.