What does 'Quality degradation' mean?

Between 50% and 90% of requests to the model in the last 15 minutes were successful. The service is running, but some users may encounter errors or timeouts. This is usually due to a temporary network load spike.

What does 'Failure' mean?

Less than half of the requests to the model were successful. The model is effectively unavailable. We open an incident and continue checking — the status is restored automatically as soon as the network responds.

Where does the data come from?

These come from our own requests to the Gonka network. We count every request that passes through the gateway: whether the response arrived from the node and the latency. No third-party monitors — we use the same nodes and the same path as user requests. The bars under each model show the percentage of successful requests hourly for the day; hours without client requests remain empty.

How often is the page updated?

The widget automatically reloads data every minute. To keep the data fresh in real time, you can simply leave the tab open.

Gonka Network

Network Status

Real-time model availability and latency. Automatically updated every minute.

Overall

Degraded performance

MiniMax M2.7

by MiniMax

Operational

Success rate, 15 min: 99%
Uptime 24h: 77.6%
Uptime 7d: 95.4%
Median latency: 306 ms
95th percentile latency: 2.36 с
Checks in 15 min: 11 / 15
Last check: 2 minutes ago
Input Context: 195K
Max Output: 8K
VRAM: 320 GB

Successful requests per hour for the last 24h

24h agonow

DeepSeek V4 Flash

by DeepSeek

Degraded

Success rate, 15 min: 54%
Uptime 24h: 78.9%
Uptime 7d: 95.0%
Median latency: 373 ms
95th percentile latency: 3.90 с
Checks in 15 min: 11 / 15
Last check: 2 minutes ago
Input Context: 371K
Max Output: 32K
VRAM: 280 GB

Successful requests per hour for the last 24h

24h agonow

Active Incidents

Degraded
DeepSeek V4 Flash
9/6/2026, 4:05:51 PM
ongoing · 17 min
Degraded
MiniMax M2.7
9/6/2026, 4:02:51 PM
ongoing · 20 min

Incidents in the last 48 hours

Degraded
DeepSeek V4 Flash
9/6/2026, 4:02:51 PM → 9/6/2026, 4:04:22 PM
2 min
Degraded
DeepSeek V4 Flash
9/6/2026, 3:44:22 PM → 9/6/2026, 3:45:22 PM
1 min
Degraded
DeepSeek V4 Flash
9/6/2026, 3:40:51 PM → 9/6/2026, 3:43:22 PM
3 min
Degraded
MiniMax M2.7
9/6/2026, 3:40:51 PM → 9/6/2026, 3:45:21 PM
5 min
Degraded
MiniMax M2.7
9/6/2026, 3:14:51 PM → 9/6/2026, 3:19:21 PM
5 min
Degraded
DeepSeek V4 Flash
9/6/2026, 3:14:51 PM → 9/6/2026, 3:20:21 PM
6 min
Degraded
DeepSeek V4 Flash
9/6/2026, 3:06:51 PM → 9/6/2026, 3:13:21 PM
7 min
Degraded
MiniMax M2.7
9/6/2026, 3:06:51 PM → 9/6/2026, 3:13:21 PM
7 min
Degraded
DeepSeek V4 Flash
9/6/2026, 3:03:51 PM → 9/6/2026, 3:05:22 PM
2 min
Degraded
MiniMax M2.7
9/6/2026, 3:03:51 PM → 9/6/2026, 3:05:22 PM
2 min
Degraded
DeepSeek V4 Flash
9/6/2026, 2:57:51 PM → 9/6/2026, 3:00:22 PM
3 min
Degraded
MiniMax M2.7
9/6/2026, 2:57:51 PM → 9/6/2026, 3:00:22 PM
3 min
Degraded
MiniMax M2.7
9/6/2026, 2:55:23 PM → 9/6/2026, 2:56:21 PM
1 min
Degraded
DeepSeek V4 Flash
9/6/2026, 2:53:51 PM → 9/6/2026, 2:55:22 PM
2 min
Degraded
MiniMax M2.7
9/6/2026, 2:53:51 PM → 9/6/2026, 2:54:21 PM
1 min
Degraded
DeepSeek V4 Flash
9/6/2026, 2:24:51 PM → 9/6/2026, 2:51:21 PM
27 min
Degraded
DeepSeek V4 Flash
9/6/2026, 2:20:51 PM → 9/6/2026, 2:22:21 PM
2 min
Degraded
MiniMax M2.7
9/6/2026, 2:20:51 PM → 9/6/2026, 2:52:22 PM
32 min
Degraded
DeepSeek V4 Flash
9/6/2026, 2:07:22 PM → 9/6/2026, 2:08:22 PM
1 min
Degraded
MiniMax M2.7
9/6/2026, 2:07:21 PM → 9/6/2026, 2:08:22 PM
1 min
Degraded
DeepSeek V4 Flash
9/6/2026, 1:42:51 PM → 9/6/2026, 2:05:21 PM
23 min
Degraded
MiniMax M2.7
9/6/2026, 1:42:51 PM → 9/6/2026, 2:05:21 PM
23 min
Degraded
DeepSeek V4 Flash
9/6/2026, 1:38:51 PM → 9/6/2026, 1:39:21 PM
1 min
Degraded
MiniMax M2.7
9/6/2026, 1:38:51 PM → 9/6/2026, 1:39:21 PM
1 min
Degraded
DeepSeek V4 Flash
9/6/2026, 1:34:51 PM → 9/6/2026, 1:37:22 PM
3 min
Degraded
MiniMax M2.7
9/6/2026, 1:34:51 PM → 9/6/2026, 1:36:21 PM
2 min
Degraded
MiniMax M2.7
9/6/2026, 1:28:22 PM → 9/6/2026, 1:30:21 PM
2 min
Degraded
MiniMax M2.7
9/6/2026, 1:24:22 PM → 9/6/2026, 1:26:21 PM
2 min
Degraded
DeepSeek V4 Flash
9/6/2026, 1:20:21 PM → 9/6/2026, 1:33:22 PM
13 min
Degraded
DeepSeek V4 Flash
9/6/2026, 1:17:51 PM → 9/6/2026, 1:19:21 PM
2 min
Degraded
DeepSeek V4 Flash
9/6/2026, 12:53:51 PM → 9/6/2026, 1:15:22 PM
22 min
Degraded
MiniMax M2.7
9/6/2026, 12:53:51 PM → 9/6/2026, 1:23:22 PM
30 min
Degraded
MiniMax M2.7
9/6/2026, 12:41:21 PM → 9/6/2026, 12:52:21 PM
11 min
Degraded
DeepSeek V4 Flash
9/6/2026, 12:36:51 PM → 9/6/2026, 12:52:22 PM
16 min
Degraded
MiniMax M2.7
9/6/2026, 12:36:51 PM → 9/6/2026, 12:40:21 PM
4 min
Degraded
MiniMax M2.7
9/6/2026, 12:33:51 PM → 9/6/2026, 12:34:21 PM
1 min
Degraded
DeepSeek V4 Flash
9/6/2026, 12:29:51 PM → 9/6/2026, 12:35:21 PM
6 min
Degraded
MiniMax M2.7
9/6/2026, 12:29:51 PM → 9/6/2026, 12:32:21 PM
3 min
Degraded
MiniMax M2.7
9/6/2026, 11:56:51 AM → 9/6/2026, 12:25:21 PM
29 min
Degraded
MiniMax M2.7
9/6/2026, 11:49:51 AM → 9/6/2026, 11:55:21 AM
6 min
Degraded
MiniMax M2.7
9/6/2026, 11:44:51 AM → 9/6/2026, 11:46:21 AM
2 min
Degraded
DeepSeek V4 Flash
9/6/2026, 11:17:51 AM → 9/6/2026, 11:19:21 AM
2 min
Degraded
MiniMax M2.7
9/6/2026, 11:10:51 AM → 9/6/2026, 11:32:22 AM
22 min
Degraded
MiniMax M2.7
9/6/2026, 11:05:21 AM → 9/6/2026, 11:08:21 AM
3 min
Degraded
DeepSeek V4 Flash
9/6/2026, 11:03:51 AM → 9/6/2026, 11:04:21 AM
1 min
Degraded
MiniMax M2.7
9/6/2026, 10:35:51 AM → 9/6/2026, 11:04:21 AM
29 min
Degraded
DeepSeek V4 Flash
9/6/2026, 10:35:51 AM → 9/6/2026, 11:01:21 AM
26 min
Degraded
DeepSeek V4 Flash
9/6/2026, 10:24:51 AM → 9/6/2026, 10:34:21 AM
10 min

Models under monitoring

MiniMax M2.7

by MiniMax. Available via the unified OpenAI- and Anthropic-compatible JoinGonka API. Current status and uptime can be found in the cards above.

DeepSeek V4 Flash

by DeepSeek. Available via the unified OpenAI- and Anthropic-compatible JoinGonka API. Current status and uptime can be found in the cards above.

Frequently asked questions

What does 'Quality degradation' mean?: Between 50% and 90% of requests to the model in the last 15 minutes were successful. The service is running, but some users may encounter errors or timeouts. This is usually due to a temporary network load spike.
What does 'Failure' mean?: Less than half of the requests to the model were successful. The model is effectively unavailable. We open an incident and continue checking — the status is restored automatically as soon as the network responds.
Where does the data come from?: These come from our own requests to the Gonka network. We count every request that passes through the gateway: whether the response arrived from the node and the latency. No third-party monitors — we use the same nodes and the same path as user requests. The bars under each model show the percentage of successful requests hourly for the day; hours without client requests remain empty.
How often is the page updated?: The widget automatically reloads data every minute. To keep the data fresh in real time, you can simply leave the tab open.

How we measure

The model status is calculated based on the percentage of successfully processed requests over the last 15 minutes: 'Operational' — 90% and above, 'Degraded' — from 50% to 90%, 'Down' — below 50%. Calculations include real client requests, and only network-side failures are counted: timeouts, node overload, or lack of available workers. Request-level errors — such as incorrect format or insufficient balance — do not affect the status. Additionally, we send a custom test request once a minute to monitor the model's state when no client requests are present.