Default footprint ranking
Coding agents ranked by the total token cost of their latest captured version — the baseline a model reads on the first turn. Click a column header to re-sort.
Footprint extremes
17.2× spread
| # | ||||||
|---|---|---|---|---|---|---|
| 1 | 21,618 | 26 | 13 | claude-opus-4-8 | v2.1.178 | |
| 2 | 18,540 | 33 | 0 | gpt-5 | v2026.6.6 | |
| 3 | 16,379 | 15 | 0 | gpt-5 | v2026.06.15-18-00-12-6f5a2cf | |
| 4 | 15,744 | 25 | 7 | gpt-5 | v0.2.53 | |
| 5 | 12,395 | 17 | 1 | gpt-5 | v1.0.62 | |
| 6 | 11,686 | 27 | 0 | gpt-5 | v0.16.0 | |
| 7 | 9,762 | 13 | 2 | gpt-5 | v7.3.46 | |
| 8 | 8,730 | 11 | 5 | gpt-5.5 | v0.140.0 | |
| 9 | 6,785 | 9 | 1 | gpt-5 | v1.17.7 | |
| 10 | 3,890 | 25 | 0 | gpt-5 | v3.0.24 | |
| 11 | 1,255 | 4 | 0 | gpt-5 | v0.79.4 |
One row per agent, using its newest version with a captured primary request.
Token counts use the fixed o200k_base tokenizer.