Attributable launch-day reports · historical record
User reports are preserved—not projected onto Pareto or a removed alias
These reports concern the R1 launch period. R3 found no current Go-alias listing and sent no requests, so none of these accounts are current measurements of paid Pareto or the former Go identifier.
Read carefully: reported 429s, slowness, loops and errors have incomplete conditions and are not an overall failure rate. R1 probes stopped at DNS; R2 and R3 made no authorized inference request (
n=0).| Author / source | Task / route / settings | Reported result | Boundary |
|---|---|---|---|
| Scott @scottstts | “Monica’s apt test”; reasoning high; route/prompt/raw trace absent. | Said it failed after >1 hour; separately reported 13 tool calls in an hour and 70–80 tps. | First-hand coding/tool report, but method incomplete; not a benchmark. S12 |
| filipe @filicroval | Client, route and prompt unknown. | Reported about an hour without visible thinking traces. | UX/availability report; no raw session. S13 |
| Tak @cherry_mx_reds | OpenRouter; provider errors; later “Spidey bench” details absent. | Reported provider errors; self-reported 70 tok/s and 81% cache hit. | Operational observation, not a quality benchmark. S14 |
| Tim Jayas @TimJayas | OpenRouter; prompt/client/settings unknown. | Reported errors every run and extreme slowness. | Cannot establish an outage. S15 |
| Reddit commenters | OpenCode/OpenRouter inconsistently mentioned; mostly no settings. | Reports include stuck thinking, no response, loops/elementary errors and timeouts; other accounts differ. | Community reports; do not aggregate. S10 |
What a comparable review needs
Task
Prompt + input
Include sensitivity, context size and expected output.
Route
Model + provider + client
Also record tools, reasoning setting, retries and fallback.
Outcome
Raw/redacted trace
UTC, latency, status/429 headers, errors and comparison control.
Current route status: current availability audit.