Measurements of the independent monitor Gonka Space (power.gnk.space), model DeepSeek-V4-Flash, tariff Super Deal. SAGG stands outside the main list in this monitor. These are measurements for the stated period, not a promise.
94.4%Availability, last 24 hours
94.4%Good answers, last 24 hours
3.22 sTime to first text, median
Median of the other providers (8): availability 93.6%, good answers 89.7%, time to first text 3.65 s.
History is accumulating from 2026-11-07, 5 day(s) collected.
Daily values
Date (UTC)
Availability: SAGG / others
Good answers: SAGG / others
First text, median: SAGG / others
2026-11-11
98.1% / 94.6%
98.1% / 89.6%
3.24 s / 2.87 s
2026-11-10
95.3% / 95.5%
95.3% / 88.9%
5.38 s / 3.89 s
2026-11-09
100.0% / 98.0%
100.0% / 93.9%
1.83 s / 4.03 s
2026-11-08
99.1% / 96.4%
99.1% / 91.3%
1.95 s / 3.13 s
2026-11-07
100.0% / 98.0%
100.0% / 90.2%
1.63 s / 3.36 s
Hourly values
Good answers, independent monitor power.gnk.space, last 7 days (higher is better). ━━ SAGG ╌╌ median of the other providers Time to first text, independent monitor power.gnk.space, last 7 days (lower is better). ━━ SAGG p50 ━━ SAGG p90 ╌╌ others p50 (median)
Daily values: the last snapshot of each UTC day (the monitor's own rolling 24 hours at the end of that day); a snapshot is taken every hour. "Others" is the median of the other active providers of DeepSeek-V4-Flash in the monitor. Good answers: the share of useful answers (usefulAnswerRate).
Second monitor: G-Meter
Probe results of the public monitor G-Meter (Gonka Labs, meter.gonka.gg) for the SAGG broker, as published, including provider-side errors such as 503. Nothing is filtered or smoothed. These are measurements for the stated period, not a promise.
9.1%Failed probes, latest quick run (1 of 11)
11.5 sLatency of the run
23 t/sOutput speed
Median of the other brokers (11, latest quick runs): failed probes 16%, latency 9.0 s.
Failed probes, G-Meter (Gonka Labs), last 7 days (lower is better). ━━ SAGG, quick runs ◇ SAGG, limits runs ╌╌ median of the other brokers Each dot is one run; diamonds are limits runs.Latency, G-Meter (Gonka Labs), quick runs, last 7 days (lower is better). ━━ SAGG ╌╌ median of the other brokers Output speed, G-Meter (Gonka Labs), quick runs, last 7 days (higher is better). ━━ SAGG SAGG only.
Latest runs
Started (UTC)
Type
Failed probes
Latency
Speed
Failed tests
2026-11-12 16:10
quick
1 of 11 (9.1%)
11.5 s
23 t/s
json_mode
2026-11-12 15:00
limits
2 of 3 (67%)
64.5 s
63 t/s
max_input, max_output
2026-11-12 14:00
quick
0 of 11 (0%)
4.8 s
4 t/s
2026-11-12 11:49
quick
0 of 11 (0%)
6.0 s
39 t/s
2026-11-12 11:39
limits
1 of 3 (33%)
153 s
49 t/s
max_input
2026-11-12 09:53
quick
0 of 11 (0%)
3.1 s
40 t/s
2026-11-12 08:42
quick
1 of 11 (9.1%)
6.1 s
44 t/s
json_mode
2026-11-12 08:02
limits
1 of 3 (33%)
302 s
10 t/s
max_output
2026-11-12 07:23
quick
0 of 11 (0%)
4.4 s
21 t/s
2026-11-12 06:11
quick
1 of 11 (9.1%)
4.0 s
24 t/s
json_mode
A G-Meter run is a set of test requests (quick: about 11 probes, limits: 3). "Others" is the median over the other brokers of their latest quick run in the window G-Meter publishes (its last 100 runs, about 7 hours), taken at each poll; we poll every 30 minutes and keep every run. Latency and speed are plotted for quick runs only; limits runs measure long requests and are not comparable. Failed probes: the share of probes that did not succeed.
Short answers. "Served" means the answer was delivered; it does not mean the answer was clean. Providers of the decentralized network require a minimum of 64 generated tokens on every request, a known behaviour of the network (gonka-ai/gonka issue 1955), so on very short requests (a one-word answer, a short JSON) they append repeated text. In our test of 2026-10-27, 17% to 23% of a mixed set of 60 requests were affected when the decentralized-network providers were called directly (15% and 18% through our two lines), up to 70% of one-word requests and up to 50% of short JSON requests. In the R1 24-hour test, 3.8% to 5.4% of the short answers served by the decentralized-network providers ended at the output limit. The provider outside the decentralized network did not show this (0 of 243 in R1, 0 of 60 on 2026-10-27). The network has published a fix (gonka-ai/gonka pull request 1972); when it will be switched on is not known. Super Deal (economical tariff, depends on the decentralized network, no insurance outside that network) has no provider outside the decentralized network, and on Standard that provider is used only when the earlier ones fail. Since 2026-10-30 the gateway cuts the repeated tail before the answer is delivered, on both lines, and bills only the delivered text. In a live check on that day (one request at a time, a unique mark in every prompt), 2.0% (0.8% to 5.0%) of 200 answers on Super Deal (economical tariff, depends on the decentralized network, no insurance outside that network) and 2.9% (1.5% to 5.5%) of 280 answers on Standard still contained a repeat or text where the model thinks aloud; 96.5% and 95.5% of the requests of a labelled set of short questions were answered right, against 91.5% (410 requests) for the provider outside the decentralized network. These are short single checks, not a promise. Details: benchmarks.