How it works

Parallel streams: what they are and how many you need

At Tenzo Cloud you pay for streams, not tokens. Here is what that means in practice and how to choose a number of streams so you neither overpay nor wait.

Tenzo Cloud teamPublished Oct 8, 20262 min read

Start freefrom $11 a month per stream · 3 hours with every model, no card

A stream is a concurrent request

A stream is how many requests the model handles for you at the same moment. With one stream the next request starts when the previous one ends. With four, four answers are generated in parallel and the fifth waits for a free slot.

A stream does not limit volume. Within a stream you can send as many requests in a row and as many tokens as you like. Only “how many at once” is limited.

How many to take

Scenario	Streams	Why
Chat for yourself	1	You ask one thing at a time
A developer in Cursor or Claude Code	1–2	The editor sometimes makes background calls
An agent with helpers	4	Helpers run in parallel
A team of 3–5	4	Requests rarely overlap
A bot with steady traffic	4–8	Peak hours, parallel customers

What happens when you run short

The extra request gets a 429 error with code concurrency_limit. Most SDKs and agents retry it automatically after a couple of seconds, so you may notice a short pause at most.

If 429s come all the time, you really need more streams. Streams for one model stack, and the volume discount starts at two.

Why streams instead of tokens

A token bill is unpredictable: an agent on a hard task can burn ten times the usual. Streams limit load, not volume, so the monthly price is known upfront and long context and long sessions cost nothing extra.

A real-life example: a support bot

Say a bot answers customers of an online store. An average answer takes 5–10 seconds. If 300 messages arrive in a peak hour, that is one every 12 seconds on average, so one stream almost suffices. But messages arrive unevenly: after a newsletter twenty may come in a minute.

With one stream the twentieth customer would wait a couple of minutes. With four streams, about half a minute. So for bots with live customers we suggest sizing for the peak, not the average.

How to tell you are short on streams

Your service logs show 429 concurrency_limit responses
The agent visibly stalls between steps
Bot users complain about slow answers at peak hours
The stream load ring in the dashboard is often full

How to split streams between services

If one subscription serves both people and automation, cap streams for the automation key. For example, give the bot at most 4 of 8 streams so staff in the chat still have free streams during a rush.

The cap is set in the key settings in the dashboard and applies immediately, no service restart needed.

Price per stream

Qwen 3.8 Max
Operational
1 stream 4+ streams, each
3 days $2.73₸1,150 $2.10₸880
7 days $5.33₸2,280 $4.10₸1,750
30 days $14.30₸6,180 $11₸4,750
GLM 5.3
Coming soon
1 stream 4+ streams, each
3 days $2.86₸1,240 $2.20₸950
7 days $5.72₸2,470 $4.40₸1,900
30 days $15.73₸6,830 $12.10₸5,250

	1 stream	4+ streams, each
3 days	$2.73₸1,150	$2.10₸880
7 days	$5.33₸2,280	$4.10₸1,750
30 days	$14.30₸6,180	$11₸4,750

	1 stream	4+ streams, each
3 days	$2.86₸1,240	$2.20₸950
7 days	$5.72₸2,470	$4.40₸1,900
30 days	$15.73₸6,830	$12.10₸5,250

Price per stream. Volume discount: 2–3 streams pay 88%, 4 or more pay 77% of the single-stream price. Streams and periods stack. Prices sync from the dashboard automatically.

FAQ

Can I split streams between keys?

Yes. Each key can have a maximum number of streams so one service cannot take them all.

Are streams shared between chat and API?

Yes, there is one subscription per model: chat and key requests use the same streams.

Is there a per-minute limit?

Yes, up to 120 new requests a minute per account, a guard against runaway loops rather than a limit for normal work.

Try it on your own work

Three free hours with every model, no card and no VPN. Connect Cursor, Claude Code or your own code in a minute.

Start free See pricing