Signals
Telegram benchmark — how fast it answers
For a high-performance website, the bar for a page response is about 250 milliseconds. Creating an output — an image, a video, a written answer — is heavier work, so we hold a different bar. We hold one standard: five seconds — and we already meet it across text, image, and video.
Speed and quality
A real Telegram round-trip, today. Quality is scored 1–5, and every output is a real file you can read or watch.
| Output | Reply | Quality |
|---|---|---|
| Text | Instant — under a second | 4 / 5 |
| Image | ~5s | 4 / 5 |
| Video (short) | ~5s | 4 / 5 |
These are worst-case figures for an all-open-source model combination — and we already sit at the five-second standard. Even the longest video stays under 25 seconds.
Two ways to ask
Inside Telegram, a request can take two forms.
| Form | How | Options |
|---|---|---|
| Just ask | Plain words — “explain X”, “make an image of…”, “make a video that…”. | Sensible defaults. The five-second standard. |
The form /form |
A tap-to-toggle panel, then Generate. | Length 5 / 10 / 20s · Quality Draft / Standard / High · Priority Fastest / Balanced / Best |
No tokens to manage, no model to pick — you ask, you get the output. The project is under constant optimization, but it already meets the standard for a pleasant experience.