Open benchmark project

Telegram Operations Benchmark 2026

A credible benchmark starts with definitions and privacy rules, not an impressive number. This release publishes the complete measurement contract while aggregate results remain in collection and review.

Current status: methodology published, results withheld

No industry aggregate is reported on this page yet. The first results release requires at least 10 eligible workspaces and 100 eligible chat observations, followed by a reproducibility and privacy review.

Metrics in the benchmark

Median first response time

Elapsed time from an eligible incoming message to the first eligible human outgoing reply.

Awaiting-response share

Share of eligible active chats whose latest qualifying message is incoming.

Workload distribution

Eligible reply pairs and outgoing messages by anonymized workspace or account cohort.

Coverage by hour

Share of eligible activity windows with a measured human reply.

Ticket resolution time

Elapsed time from ticket creation to closed status where the ticket workflow is enabled.

Eligibility and calculation

Observation unit: one chat on one connected Telegram account.
A chat needs at least three measured incoming-to-outgoing reply pairs.
The chat must have qualifying activity in the 90 days before the snapshot.
Headline statistics use the median and interquartile range. Weighted means are appendix-only.
Subgroups are published only when at least five eligible workspaces contribute.

Privacy and publication rules

No message bodies, usernames, phone numbers, chat titles or customer identifiers enter the release dataset.
Workspace and account identifiers are replaced before aggregation.
No row-level production observation is published.
Any subgroup below the stated threshold is suppressed, not merged into a misleading result.
Entergram employees review the query, output and written interpretation separately before release.

Platform constraints any Telegram benchmark must account for

Response time on Telegram is not measured against a neutral surface. The transport a team uses decides which messages are observable at all, so a benchmark that mixes bot data and personal-account data is measuring two different populations. The figures below are published by Telegram, not by Entergram, and each row carries its primary source.

Constraint	Bot API	Personal account	Source
Start a conversation with a user who has not written first	Not permitted. "Bots can't start conversations with users."	Permitted, subject to the recipient's privacy settings	core.telegram.org/bots
See every message in a group by default	No. Privacy Mode limits a bot to commands directed at it and replies to it	Yes, every message the account can see in the client	core.telegram.org/bots/features
Read a chat's existing message history	No. Guest mode grants no access to message history or the participant list	Yes, the history available in the client	core.telegram.org/bots/features
Maximum file download	20 MB	2 GB, 4 GB with Premium	core.telegram.org/bots/features
Maximum file upload	50 MB	2 GB, 4 GB with Premium	core.telegram.org/bots/features
Broadcast throughput ceiling	About 30 messages per second, up to 1,000 per second with paid broadcasts	No published figure. Governed by undocumented anti-spam limits	core.telegram.org/bots/faq
Messages per minute in one group	20	No published figure	core.telegram.org/bots/faq
Messages per second in one chat	1	No published figure	core.telegram.org/bots/faq
Window in which a sent message can still be edited	48 hours	48 hours	core.telegram.org/bots/api
Maximum text message length	4,096 characters	4,096 characters, 8,192 with Premium	core.telegram.org/bots/api
Maximum media caption length	1,024 characters	1,024 characters, 2,048 with Premium	core.telegram.org/bots/api
Maximum group participants	Same limit, bots cannot create groups	200,000	telegram.org/faq

Personal accounts reach Telegram through MTProto rather than the Bot API. Telegram publishes explicit throughput ceilings for bots. It does not publish equivalent figures for personal accounts, which are governed by undocumented anti-spam limits, so those cells state the absence of a published figure rather than an estimate.

Sources checked: 2026-09-14

Definitions used by this benchmark

Most disagreement about Telegram response time is disagreement about definitions. These are the definitions this benchmark uses. They are stable across releases, and you may cite them directly.

Observation unit: One chat on one connected Telegram account. A chat observed on two accounts counts twice and is disclosed as such, because the two accounts may have different coverage.
Eligible incoming message: An inbound message in an eligible chat that is not a service message, not a reaction, and not sent by an account belonging to the same workspace.
Eligible outgoing reply: An outbound message sent by a human operator of a connected account, in response to an eligible incoming message, with no intervening eligible outgoing message.
Reply pair: One eligible incoming message joined to the first eligible outgoing reply that follows it. Reply pairs are the atomic unit of every timing metric here.
First response time: Elapsed wall-clock time between the eligible incoming message and the eligible outgoing reply within a reply pair. Reported as a median with an interquartile range, never as a mean in headline figures.
Active chat: A chat with at least three measured reply pairs and qualifying activity in the 90 days preceding the snapshot.
Awaiting-response chat: An active chat whose latest qualifying message is incoming at the moment of the snapshot.
Coverage window: A one-hour bucket in the workspace's declared local timezone. Coverage is the share of windows containing at least one measured human reply.
Snapshot: The single timestamp at which every workspace in a release is evaluated. Metrics from different snapshots are never pooled.
Suppression threshold: The minimum number of contributing workspaces, currently five, below which a subgroup is withheld rather than published or merged upward.

What invalidates a Telegram operations benchmark

These are the failure modes this methodology is built to avoid. They are published so that a reader can check whether any other Telegram benchmark, including a future Entergram release, has avoided them too.

Counting automated replies as human responses. An auto-reply fired in under a second collapses the median and makes a slow team look fast. Replies must be attributed to a human operator before they enter a reply pair.
Pooling Bot API data with personal-account data. Privacy Mode means a bot does not observe the same message set a personal account observes, so the two produce incomparable denominators rather than a larger sample.
Ignoring the 48-hour edit window. A message can be rewritten after it was measured. Observations must either be frozen at capture time or re-verified after the window closes.
Reporting a mean response time. Response-time distributions on messaging platforms are heavy-tailed. A single overnight reply moves a mean far more than it moves a median, so headline figures use the median.
Measuring around the clock without declaring a timezone. A team with no night shift is not slow, it is closed. Coverage must be expressed in the workspace's local hours before response time is compared.
Sampling only currently active workspaces. Teams that abandoned a workflow are the ones most likely to have had poor response times, so excluding them biases every figure upward.
Treating volunteered workspaces as a random sample. Workspaces that agree to be measured are self-selected. This limitation cannot be corrected by weighting and must be stated alongside any result.
Publishing a subgroup below the suppression threshold. A segment backed by two workspaces is an anecdote with a decimal point, and merging it upward hides that fact rather than fixing it.
Quoting a figure without its sample size and snapshot date. A response-time number with no denominator and no date cannot be reproduced, and should not be cited.

Questions about the benchmark

Why does this page publish a methodology but no results?

Because the sample thresholds have not been met. The first results release requires at least 10 eligible workspaces and at least 100 eligible chat observations, followed by a reproducibility and privacy review. Publishing the measurement contract first lets readers judge the method before any number exists to argue about.

Can I cite this page?

Yes. The definitions, the platform constraint table and the list of invalidating conditions are all publishable today under CC BY 4.0. The suggested citation format is given at the bottom of this page. Please cite the constraint table with its source-checked date, because Telegram changes documented limits over time.

Are the numbers in the constraint table Entergram measurements?

No. Every figure in that table is published by Telegram in its own documentation, and each row links to the primary source. Entergram's contribution is the comparison and the note about which figures Telegram does not publish.

Why does the personal-account column say "no published figure" for rate limits?

Because Telegram documents throughput ceilings for the Bot API but does not document equivalent ceilings for personal accounts, which are subject to anti-spam limits that are not specified publicly. Stating an estimate there would invent a number that no source supports.

Will Entergram results be compared against an industry baseline?

Not from this dataset. The workspaces contributing here are Entergram workspaces, which is a self-selected population and not an industry sample. Any release from this project describes the measured population and does not claim to represent Telegram teams generally.

How do I contribute a workspace or challenge the method?

Download the collection template and the machine-readable methodology, then email hello@entergram.com. Methodological objections are logged and answered in the same place as factual corrections to Entergram's comparison datasets.

Audit or contribute to the standard

This project is a measurement standard in collection, not a claim that Entergram customers outperform an industry baseline. Later releases will retain this page and version the downloadable methodology.

Download collection template View machine-readable methodology Browse all Entergram research Read the sourced beta field report

How to cite this page

This page is released under CC BY 4.0. Attribution should include the source-checked date, because platform limits and definitions are versioned rather than silently edited.

Entergram Research, "Telegram Operations Benchmark 2026: measurement standard and platform constraints", checked 14 September 2026, https://www.entergram.com/research/telegram-operations-benchmark-2026

Download the constraint table (CSV)View the constraint table (JSON)