Avant

Showing posts with label project-maya. Show all posts
Showing posts with label project-maya. Show all posts

Monday, November 9, 2026

Deepseek-v4.1-Flash-1017 benchmark on 3x RTX 5060 Ti 16GB with Dual Intel Xeon E5-2683 v4

Using Project Maya (https://github.com/mw00/project-maya) using deepseek-maya-xs With 1M context


llama-benchy --base-url http://localhost:8080/v1 --model deepseek-v4.1-flash --depth 0 --pp 2048 --tg 512 --concurrency 1 --runs 3 --latency-mode generation --no-cache --save-total-throughput-timeseries

| model               |   test |                 t/s |     peak t/s |    ttfr (ms) |   est_ppt (ms) |     e2e_ttft (ms) |
|:--------------------|-------:|--------------------:|-------------:|-------------:|---------------:|------------------:|
| deepseek-v4.1-flash | pp2048 | 161102.16 ± 9350.12 |              | 16.45 ± 0.88 |   11.82 ± 0.88 | 11712.44 ± 247.77 |
| deepseek-v4.1-flash |  tg512 |        29.56 ± 1.54 | 30.33 ± 1.70 |              |                |                   |

llama-benchy --base-url http://localhost:8080/v1 --model deepseek-v4.1-flash --depth 0 --pp 2048 --tg 512 --concurrency 4 --runs 3 --latency-mode generation --no-cache --save-total-throughput-timeseries

| model               |        test |   t/s (total) |           t/s (req) |     peak t/s |   peak t/s (req) |     ttfr (ms) |   est_ppt (ms) |       e2e_ttft (ms) |
|:--------------------|------------:|--------------:|--------------------:|-------------:|-----------------:|--------------:|---------------:|--------------------:|
| deepseek-v4.1-flash | pp2048 (c4) |  88.81 ± 1.43 | 47116.81 ± 14605.95 |              |                  | 46.70 ± 12.56 |  42.59 ± 12.56 | 47575.89 ± 26684.74 |
| deepseek-v4.1-flash |  tg512 (c4) |  18.29 ± 0.67 |        31.79 ± 2.47 | 34.33 ± 0.94 |     32.17 ± 2.58 |               |                |                     |


llama-benchy --base-url http://localhost:8080/v1 --model deepseek-v4.1-flash --depth 0 --pp 2048 --tg 512 --concurrency 8 --runs 3 --latency-mode generation --no-cache --save-total-throughput-timeseries

| model               |        test |   t/s (total) |          t/s (req) |     peak t/s |   peak t/s (req) |     ttfr (ms) |   est_ppt (ms) |       e2e_ttft (ms) |
|:--------------------|------------:|--------------:|-------------------:|-------------:|-----------------:|--------------:|---------------:|--------------------:|
| deepseek-v4.1-flash | pp2048 (c8) |  80.61 ± 0.93 | 29333.59 ± 9435.68 |              |                  | 72.39 ± 18.65 |  68.07 ± 18.65 | 96932.54 ± 55296.54 |
| deepseek-v4.1-flash |  tg512 (c8) |  17.73 ± 0.53 |       31.21 ± 1.73 | 34.00 ± 0.82 |     31.71 ± 1.72 |               |                |                     |