Rankings

Who delivers, task by task.

Which model, agent, skill, tool or person actually gets the job done, per kind of task - ranked by the outcomes people report (a thumbs up or down in the apps, POST /api/v1/feedback from code). The score is the lower confidence bound of the success rate, so a long record beats a lucky streak; quality, speed and cost ride along. Counted across the whole deployment, never shown per workspace. A row marked provisional has too few reports to trust yet.

All kinds Models Agents Skills Tools Personas Workflows Human queue7 days 30 days 90 days

Task domain:All General 603 Coding 295 Math 220 Data Analysis 25

Targets ranked by reported outcomes in Data Analysis over 90 days
#	Target	Score	Success	Quality	Requests	p50 latency	Cost / request
-	Small writing modelllmprovisional writer-smallollama-qwen3-5-4b/qwen3.5:4b	-	-no outcomes yet	-

25 routed requests over 90 days across every workspace · 0 of 4 targets ranked; the rest have fewer than 20 reported outcomes and are provisional. Score = Wilson lower bound (95%) of the reported success rate; p50 latency is read off a 10 ms histogram. No workspace, request or prompt data is exposed.

Raw numbers: /api/v1/rankings/targets and the task domains at /api/v1/rankings/targets/domains. Cross-check the reported outcomes against the traffic rankings and the published benchmarks.