# Time-to-First-Token (TTFT) Latency on Controlled synthetic corpus Config B 1.0 (test)

Source: [Time-to-First-Token (TTFT) Latency on Controlled synthetic corpus Config B 1.0 (test)](<https://www.sota2.com/research/sota/time-to-first-token-ttft-latency-on-controlled-synthetic-corpus-config-b-1-0-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-19T07:23:12.298Z

Evaluation of the Time-to-First-Token (TTFT) latency performance using the Controlled synthetic corpus, specifically Config B version 1.0 (test set).

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Time-to-First-Token (TTFT) Latency |
| Dataset (including any stated split) | Controlled synthetic corpus Config B 1.0 (test) |
| Metric | TTFT p50 |
| Unit | TIME |
| Direction | Lower is better |
| Results loaded | 4 of 4 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [TTFT p50](<https://www.sota2.com/research/sota/time-to-first-token-ttft-latency-on-controlled-synthetic-corpus-config-b-1-0-test.md?metric=TTFT%20p50>): Lower is better; unit: TIME.

- [TTFT p95](<https://www.sota2.com/research/sota/time-to-first-token-ttft-latency-on-controlled-synthetic-corpus-config-b-1-0-test.md?metric=TTFT%20p95>): Lower is better; unit: TIME.

- [TTFT Mean Latency](<https://www.sota2.com/research/sota/time-to-first-token-ttft-latency-on-controlled-synthetic-corpus-config-b-1-0-test.md?metric=TTFT%20Mean%20Latency>): Lower is better; unit: TIME.

- [TTFT Fast Sample %](<https://www.sota2.com/research/sota/time-to-first-token-ttft-latency-on-controlled-synthetic-corpus-config-b-1-0-test.md?metric=TTFT%20Fast%20Sample%20%25>): Higher is better; unit: PERCENT.

## Recorded results

| Method | TTFT p50 | Result date | Report paper |
| --- | --- | --- | --- |
| APC+Optimized | 40.1 | Not recorded | 6a34ccf2206542251d7c5e33 |
| APC+Retrieval | 59.6 | Not recorded | 6a34ccf2206542251d7c5e33 |
| APC+Sorted | 81.2 | Not recorded | 6a34ccf2206542251d7c5e33 |
| No Cache | 121.3 | Not recorded | 6a34ccf2206542251d7c5e33 |

### APC+Optimized

TTFT p50: 40.1. Result ID: 6a34edf5fbc373592c31d046.

[Reporting paper](<https://www.sota2.com/research/paper/6a34ccf2206542251d7c5e33.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a34ccf2206542251d7c5e33.md>)

Recorded evaluation configuration: {"Hardware":"RTX 4090","Model":"Qwen2.5-7B","Strategy":"APC with optimized ordering (CacheWeaver)"}. This record does not assert independent reproduction.

### APC+Retrieval

TTFT p50: 59.6. Result ID: 6a34edf5fbc373592c31d044.

[Reporting paper](<https://www.sota2.com/research/paper/6a34ccf2206542251d7c5e33.md>)

Recorded evaluation configuration: {"Hardware":"RTX 4090","Model":"Qwen2.5-7B","Strategy":"APC with retrieval order"}. This record does not assert independent reproduction.

### APC+Sorted

TTFT p50: 81.2. Result ID: 6a34edf5fbc373592c31d045.

[Reporting paper](<https://www.sota2.com/research/paper/6a34ccf2206542251d7c5e33.md>)

Recorded evaluation configuration: {"Hardware":"RTX 4090","Model":"Qwen2.5-7B","Strategy":"APC with lexicographic sorting"}. This record does not assert independent reproduction.

### No Cache

TTFT p50: 121.3. Result ID: 6a34edf5fbc373592c31d043.

[Reporting paper](<https://www.sota2.com/research/paper/6a34ccf2206542251d7c5e33.md>)

Recorded evaluation configuration: {"Hardware":"RTX 4090","Model":"Qwen2.5-7B","Strategy":"Baseline"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.