# reason-ar-text: 431 prompts answered by every arm (1296 calls, 1 failed), arms off, low, default


### All prompts

| group | prompts | arm | first answer token p50 / p90 / p95 s | total p50 / p90 / p95 s | first word after 2 s / 3 s / 5 s | reasoning shown first p50 s | output tokens mean |
|---|---:|---|---:|---:|---:|---:|---:|
| all | 431 | off | 1.17 / 1.84 / 2.23 | 1.81 / 2.68 / 3.02 | 8% / 1% / 0% | — | 95 |
| all | 431 | low | 2.49 / 4.75 / 5.76 | 2.96 / 5.39 / 6.31 | 69% / 38% / 8% | 1.20 | 405 |
| all | 431 | default | 3.69 / 6.32 / 8.04 | 4.05 / 6.88 / 8.77 | 92% / 66% / 24% | 1.19 | 660 |

### By difficulty

| group | prompts | arm | first answer token p50 / p90 / p95 s | total p50 / p90 / p95 s | first word after 2 s / 3 s / 5 s | reasoning shown first p50 s | output tokens mean |
|---|---:|---|---:|---:|---:|---:|---:|
| easy | 144 | off | 1.14 / 1.88 / 2.09 | 1.72 / 2.49 / 2.71 | 8% / 1% / 0% | — | 89 |
| easy | 144 | low | 2.01 / 3.93 / 4.50 | 2.46 / 4.26 / 5.26 | 50% / 22% / 3% | 1.16 | 297 |
| easy | 144 | default | 2.99 / 4.83 / 5.67 | 3.41 / 5.31 / 6.34 | 83% / 49% / 8% | 1.16 | 473 |
| normal | 144 | off | 1.12 / 1.66 / 1.94 | 1.78 / 2.67 / 2.94 | 4% / 0% / 0% | — | 90 |
| normal | 144 | low | 2.65 / 4.29 / 5.25 | 3.06 / 4.95 / 5.68 | 73% / 40% / 7% | 1.15 | 393 |
| normal | 144 | default | 3.87 / 6.47 / 7.77 | 4.23 / 7.02 / 8.77 | 97% / 69% / 29% | 1.20 | 667 |
| hard | 143 | off | 1.24 / 2.10 / 2.54 | 1.99 / 2.83 / 3.52 | 11% / 3% / 1% | — | 106 |
| hard | 143 | low | 3.12 / 5.98 / 7.24 | 3.57 / 6.31 / 7.34 | 85% / 52% / 15% | 1.25 | 525 |
| hard | 143 | default | 4.46 / 7.80 / 9.84 | 4.90 / 8.29 / 10.45 | 96% / 80% / 35% | 1.24 | 841 |

### By mode

| group | prompts | arm | first answer token p50 / p90 / p95 s | total p50 / p90 / p95 s | first word after 2 s / 3 s / 5 s | reasoning shown first p50 s | output tokens mean |
|---|---:|---|---:|---:|---:|---:|---:|
| general | 36 | off | 1.18 / 1.99 / 2.54 | 1.76 / 2.65 / 2.83 | 8% / 0% / 0% | — | 70 |
| general | 36 | low | 2.27 / 4.03 / 6.14 | 2.62 / 4.33 / 6.49 | 61% / 19% / 6% | 1.14 | 334 |
| general | 36 | default | 3.32 / 8.38 / 9.54 | 3.66 / 8.77 / 9.74 | 83% / 53% / 22% | 1.17 | 632 |
| sales | 36 | off | 1.18 / 1.81 / 2.25 | 1.80 / 2.71 / 2.98 | 8% / 0% / 0% | — | 77 |
| sales | 36 | low | 1.97 / 4.27 / 5.25 | 2.46 / 4.70 / 5.68 | 47% / 28% / 6% | 1.07 | 334 |
| sales | 36 | default | 3.30 / 5.29 / 5.47 | 3.70 / 5.73 / 6.09 | 92% / 67% / 14% | 1.24 | 577 |
| recruiting | 36 | off | 1.25 / 2.08 / 2.62 | 1.72 / 2.61 / 3.12 | 14% / 3% / 0% | — | 71 |
| recruiting | 36 | low | 3.12 / 5.70 / 6.36 | 3.32 / 6.42 / 6.92 | 81% / 50% / 17% | 1.34 | 452 |
| recruiting | 36 | default | 4.53 / 6.79 / 7.80 | 4.73 / 6.92 / 8.29 | 97% / 78% / 36% | 1.25 | 779 |
| team-meet | 36 | off | 1.16 / 1.67 / 2.41 | 1.57 / 2.58 / 2.90 | 8% / 0% / 0% | — | 64 |
| team-meet | 36 | low | 2.47 / 4.33 / 5.12 | 2.89 / 4.54 / 5.64 | 78% / 28% / 6% | 1.28 | 336 |
| team-meet | 36 | default | 3.35 / 6.60 / 7.72 | 3.66 / 7.03 / 8.01 | 94% / 67% / 14% | 1.21 | 580 |
| looking-for-work | 36 | off | 1.19 / 2.00 / 2.90 | 1.81 / 2.74 / 3.52 | 11% / 3% / 0% | — | 76 |
| looking-for-work | 36 | low | 3.43 / 5.70 / 5.98 | 3.71 / 6.05 / 6.31 | 75% / 58% / 17% | 1.19 | 469 |
| looking-for-work | 36 | default | 4.68 / 9.94 / 12.89 | 4.98 / 10.26 / 13.27 | 97% / 72% / 44% | 1.23 | 795 |
| lecture | 89 | off | 1.14 / 1.83 / 2.30 | 1.94 / 2.67 / 3.16 | 7% / 1% / 0% | — | 108 |
| lecture | 89 | low | 2.93 / 4.81 / 5.53 | 3.41 / 5.43 / 6.21 | 72% / 48% / 6% | 1.20 | 447 |
| lecture | 89 | default | 3.67 / 5.55 / 6.27 | 4.07 / 5.97 / 6.75 | 93% / 67% / 21% | 1.22 | 604 |
| technical-interview | 90 | off | 1.17 / 1.64 / 1.94 | 1.86 / 2.78 / 2.91 | 4% / 0% / 0% | — | 137 |
| technical-interview | 90 | low | 2.49 / 5.14 / 7.44 | 3.06 / 5.78 / 7.70 | 68% / 34% / 10% | 1.20 | 453 |
| technical-interview | 90 | default | 3.74 / 7.43 / 9.01 | 4.12 / 8.00 / 9.62 | 91% / 63% / 24% | 1.16 | 733 |
| seminar | 36 | off | 1.08 / 1.76 / 9.16 | 1.78 / 2.58 / 9.18 | 8% / 6% / 6% | — | 105 |
| seminar | 36 | low | 2.24 / 3.97 / 6.57 | 2.83 / 5.69 / 7.13 | 58% / 28% / 8% | 1.10 | 351 |
| seminar | 36 | default | 3.50 / 6.47 / 14.70 | 4.05 / 7.32 / 15.35 | 78% / 56% / 25% | 1.13 | 648 |
| call-center | 36 | off | 1.18 / 1.92 / 2.96 | 1.74 / 2.43 / 3.40 | 6% / 3% / 0% | — | 63 |
| call-center | 36 | low | 2.61 / 3.71 / 4.38 | 2.84 / 4.17 / 4.59 | 83% / 33% / 3% | 1.24 | 332 |
| call-center | 36 | default | 3.74 / 5.25 / 6.31 | 4.03 / 5.44 / 7.32 | 97% / 75% / 19% | 1.25 | 562 |

### By how the question arrives

| group | prompts | arm | first answer token p50 / p90 / p95 s | total p50 / p90 / p95 s | first word after 2 s / 3 s / 5 s | reasoning shown first p50 s | output tokens mean |
|---|---:|---|---:|---:|---:|---:|---:|
| heard | 275 | off | 1.17 / 1.95 / 2.40 | 1.80 / 2.70 / 3.12 | 9% / 2% / 1% | — | 88 |
| heard | 275 | low | 2.54 / 4.84 / 5.74 | 2.96 / 5.39 / 6.31 | 73% / 37% / 8% | 1.20 | 392 |
| heard | 275 | default | 3.76 / 6.32 / 8.49 | 4.05 / 6.79 / 9.01 | 92% / 68% / 23% | 1.21 | 645 |
| typed | 156 | off | 1.13 / 1.81 / 2.10 | 1.82 / 2.68 / 3.01 | 5% / 1% / 0% | — | 107 |
| typed | 156 | low | 2.46 / 4.68 / 6.21 | 2.92 / 5.44 / 6.62 | 63% / 39% / 8% | 1.20 | 427 |
| typed | 156 | default | 3.50 / 6.42 / 8.04 | 4.09 / 7.02 / 8.77 | 90% / 62% / 26% | 1.16 | 686 |

### By prompt size (input tokens)

| group | prompts | arm | first answer token p50 / p90 / p95 s | total p50 / p90 / p95 s | first word after 2 s / 3 s / 5 s | reasoning shown first p50 s | output tokens mean |
|---|---:|---|---:|---:|---:|---:|---:|
| under 3k | 365 | off | 1.17 / 1.92 / 2.38 | 1.81 / 2.71 / 3.06 | 9% / 2% / 1% | — | 93 |
| under 3k | 353 | low | 2.46 / 4.63 / 5.70 | 2.90 / 5.20 / 6.12 | 69% / 35% / 7% | 1.19 | 394 |
| under 3k | 357 | default | 3.65 / 6.27 / 7.72 | 4.01 / 6.65 / 8.02 | 92% / 66% / 22% | 1.19 | 629 |
| 3k–8k | 66 | off | 1.14 / 1.55 / 1.83 | 1.79 / 2.57 / 2.71 | 0% / 0% / 0% | — | 103 |
| 3k–8k | 78 | low | 3.04 / 5.52 / 6.36 | 3.40 / 6.22 / 6.92 | 71% / 50% / 13% | 1.24 | 455 |
| 3k–8k | 74 | default | 3.89 / 7.43 / 12.89 | 4.35 / 8.00 / 13.27 | 92% / 68% / 32% | 1.20 | 808 |

### The same prompt, with reasoning minus without (paired)

| arm | prompts | first answer token later by p50 / p90 / p95 s | total later by p50 / p90 s | output tokens × |
|---|---:|---:|---:|---:|
| low | 431 | 1.31 / 3.59 / 4.82 | 1.15 / 3.45 | 4.3 |
| default | 431 | 2.42 / 5.17 / 6.81 | 2.15 / 4.90 | 6.9 |

Failed calls: {"default: 200 no answer text in the stream":1}
