Community Stats

What the herd is actually running.

These numbers come from installs that opted in to anonymous usage reporting. Every field is listed on the telemetry page. We publish them back because the community generated them.

Covering 2026-09-02 to 2026-10-02. Self-reported by participating installs, so treat them as a good signal rather than a census. The two herd counts differ on purpose: an idle herd reports its machines without routing any requests.

Herds that routed requests12
Herds reporting in16
Machines in those herds30
Requests routed260,020
Prompt tokens334,537,135
Completion tokens59,880,197

How many machines are in a herd

How many machines are in a herd Number of herds by how many machines they reported. 1 machine: 7 herds. 2 machines: 5 herds. 3 machines: 3 herds. 4 machines: 1 herd. 5+ machines: 0 herds. 1 machine: 7 herds 1 machine 7 2 machines: 5 herds 2 machines 5 3 machines: 3 herds 3 machines 3 4 machines: 1 herd 4 machines 1 5+ machines: 0 herds 5+ machines 0
Herds by machine count, each machine counted once from its latest report in this window. Self-reported by herds with telemetry on; not verified.

Models the most herds run

Counted by herds, not requests. One busy herd can send most of the traffic in the request table below, so this is the better read on what is common.

Models by number of herds running them Number of herds that routed at least one request to each model. deepseek-coder-v2:16b: 2 herds. gemma3:27b: 2 herds. nomic-embed-text: 2 herds. codestral:22b: 1 herd. command-r: 1 herd. deepseek-coder-v2-8k: 1 herd. definitely-not-a-real-model:99b: 1 herd. gemma3:12b: 1 herd. gemma3:4b: 1 herd. gemma4: 1 herd. deepseek-coder-v2:16b: 2 herds deepseek-coder-v2:16b 2 gemma3:27b: 2 herds gemma3:27b 2 nomic-embed-text: 2 herds nomic-embed-text 2 codestral:22b: 1 herd codestral:22b 1 command-r: 1 herd command-r 1 deepseek-coder-v2-8k: 1 herd deepseek-coder-v2-8k 1 definitely-not-a-real-model:99b: 1 herd definitely-not-a-real-model:99b 1 gemma3:12b: 1 herd gemma3:12b 1 gemma3:4b: 1 herd gemma3:4b 1 gemma4: 1 herd gemma4 1
Top 10 models by herds that routed at least one request to them, with "model:latest" counted as "model". Self-reported by herds with telemetry on; not verified.

Most-used models

#ModelRequestsShare
1gpt-oss:120b224,78086.45%
2nomic-embed-text21,2238.16%
3gemma3:27b11,2824.34%
4gemma4:12b11,1160.43%
5ornith-1.5:35b4340.17%
6qwen3.8:27b-mlx2860.11%
7llama3.1:8b2420.09%
8gemma4:12b1320.05%
9command-r1170.04%
10codestral:22b1050.04%
11gemma4-e4b-coder550.02%
12deepseek-coder-v2:16b530.02%
13deepseek-coder-v2-8k450.02%
14nemo-syslog330.01%
15mannix/qwen3.6-27b-a3b-coder:vision-Q3_K_M270.01%
16qwen3.5:4b230.01%
17qwen2.5:0.5b200.01%
18Qwen3_Reasoning_ToolsCalling_8B:Q8_40Kctx_4Kemb110.0%
19qwen3:32b90.0%
20gemma480.0%
21nemotron3:33b60.0%
22omlx:gpt-oss-20b-MXFP4-Q830.0%
23qwen3:14b20.0%
24gemma3:4b20.0%
25marco-mini20.0%

The herd leaderboard

Herds that report in, by name where the operator picked one and by a generated handle otherwise. Ranked by total memory, because that is more fun to argue about than request counts.

#HerdMachinesTotal memory
1 LiveNeon Geeks in the Woods 1 512 GB
2 herd-66cfd4 4 268 GB
3 herd-c62489 3 192 GB
4 herd-f7c6e4 3 140 GB
5 herd-6c30ce 2 128 GB

See the whole leaderboard →

Most common errors

#CategoryCountShare
1bad_request31499.05%
2other30.95%

Error categories are a fixed list. Raw error text never leaves anyone's machine.

Machine-readable

The same data is available as JSON at /api/v1/stats.