DrivenBench 得分
能力得分与基准测试总成本上限的关系（对数坐标）。实线为观测到的帕累托前沿。
排行榜
| 排名 | 模型 | 得分能力得分 | 能力通过数 | 成本上限 | 延迟中位数 / p90 | 详情 |
|---|---|---|---|---|---|---|
| =1=1 | 93.9%93.9% | 155/165 | $57.69 | 30s/70s | ||
| =1=1 | 93.9%93.9% | 155/165 | $59.42 | 59s/169s | ||
| 33 | 90.9%90.9% | 150/165 | $160.37 | 40s/90s | ||
| 44 | 89.7%89.7% | 148/165 | $30.64 | 24s/61s | ||
| 55 | 89.1%89.1% | 147/165 | $73.65 | 75s/158s | ||
| 66 | 87.3%87.3% | 144/165 | $25.64 | 23s/62s | ||
| 77 | 85.5%85.5% | 141/165 | $92.72 | 27s/61s | ||
| 88 | 81.8%81.8% | 135/165 | $13.74 | 23s/44s | ||
| 99 | 79.4%79.4% | 131/165 | $3.63 | 19s/39s | ||
| 1010 | 77.6%77.6% | 128/165 | $2.97 | 20s/46s | ||
| 1111 | 76.4%76.4% | 126/165 | $33.54 | 16s/32s |
选择适合你投资工作流的模型，然后在 Driven 中让它开始工作。
免费使用 Driven