Resources · n8n integration

n8n MiniMax Chat Model nodeConfigure MiniMax Chat Model in n8n.

The n8n MiniMax Chat Model node gives a root node its brain. It is a sub-node, meaning a node that never runs on its own: it plugs into the Model port of an agent or a chain. It exposes 7 parameters, from the model list down to Top P, and answers to whatever the root node asks it.

Try the MiniMax Chat Model node on n8nSee the 7 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n MiniMax Chat Model node actually do?

It supplies a MiniMax chat model to a root node. n8n splits its AI nodes in two: a root node (an AI Agent, a chain, a vector store in root mode) receives the workflow items through its main input, and sub-nodes attach to its ports, one kind per port. This node is one of those sub-nodes. It has no main input, it produces no items of its own, and it only wakes up when the root node above it needs a completion.

Picture a support inbox. An AI Agent reads an incoming message, decides whether it can answer, and drafts a reply. The agent handles the logic; the MiniMax Chat Model node attached to its Model port is what writes the sentences. Swap that sub-node for an OpenAI or Anthropic one and the rest of the workflow does not move.

Second scenario, classification at volume. Rows arrive from a spreadsheet import, a Text Classifier sorts them, and the model behind it decides the label. Here you want Sampling Temperature near zero so the same row always gets the same label, and Max Retries set so one failed call does not kill the batch.

Third, long-form drafting. A Summarization Chain feeds a transcript to the model and asks for a digest. Maximum Number of Tokens sets how long the answer is allowed to run, and Timeout decides how patient n8n is before it gives up on the request.

When should you pick something else? If the task is a single prompt with a fixed shape and no tool calling, a Basic LLM Chain with this node under it is lighter than a full agent. And if MiniMax has a feature its OpenAI-compatible surface does not expose, this node will not reach it: the node exposes what the catalog lists, nothing more.

The limits worth knowing before you build: this node holds no memory and no tools of its own, because in n8n those are separate sub-nodes on separate ports of the root node. The model list is fixed to the values the node ships with, so a model released after your n8n version will not appear. And the token ceiling depends on which model you picked, so a value that works on one entry in the list can be refused by another.

Connect

How do you authenticate the node?

  1. 01

    Get a MiniMax API key

    The node talks to MiniMax through its OpenAI-compatible text API, so it needs an account on the MiniMax platform and a key issued from there. On n8n Cloud there is a second route: pick Use Gateway credits in the node's credential field and the node runs on Gateway credits, with no MiniMax account at all. On a self-hosted instance, the key is the only way in.

  2. 02

    Create the credential once in n8n

    Open the Credentials menu, create the MiniMax credential, paste the key, and save it. A credential in n8n is a stored set of secrets, kept apart from the workflow itself. Create it once and every workflow on the instance can reuse it, which matters the day the key rotates: you edit one record instead of hunting through nodes.

  3. 03

    Attach the node to a root node

    Drop the node on the canvas and drag its output onto the Model port of the root node, an AI Agent or a chain. The connection is the point: without it the node has nothing to answer to and n8n has no reason to execute it. Once the wire is drawn, pick a value in Model and run the root node to confirm the pair works.

Connections

What do you plug this node into?

A root node runs the workflow items; sub-nodes hang off its ports and do one job each. This one occupies the Model port.

Output (what it plugs into)

  • Modelai_languageModel
01ModelThis is the output side of the node: it connects to the ai_languageModel port of a root node, which is the slot where that node looks for the language model it will call.

This is the output side of the node: it connects to the ai_languageModel port of a root node, which is the slot where that node looks for the language model it will call.

Key parameters

  • Required: the node exists only to fill this port, so an unconnected MiniMax Chat Model node sits on the canvas and never executes.
  • AI Agent: the usual root node, when the model has to pick between several actions.
  • Basic LLM Chain: one prompt in, one answer out, nothing else to configure.
  • Text Classifier: the model returns a label instead of free text.
  • Summarization Chain: long input, condensed output, same port.
Use cases
one MiniMax Chat Model node per root node, rewired when a prompt moves from a chain to a full agent.
Parameters

Which parameters matter, and when?

The MiniMax Chat Model node has 7 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.

01

Model

model

What you see in n8n

Notes & use cases

Picks which MiniMax model generates the completion. Everything else on this node tunes the call; this one decides who answers it.

Key parameters

  • Model: the dropdown lists MiniMax-M2, MiniMax-M2.1, MiniMax-M2.1-highspeed, MiniMax-M2.5, MiniMax-M2.5-highspeed, MiniMax-M2.7 and MiniMax-M2.7-highspeed.
Use cases
set it from an expression such as {{ $json.model }} when a routing step upstream decides which entry to use per item, rather than duplicating the whole branch.
02

Hide Thinking

options.hideThinking

What you see in n8n

Notes & use cases

Strips the chain-of-thought reasoning out of the response so the root node receives the final answer alone. It is on by default, and the node removes the <think> tags before passing anything up.

Key parameters

  • Hide Thinking: a switch. Turned off, the reasoning stays in the output and travels downstream with the answer.
Use cases
turn it off while debugging a prompt that keeps landing on the wrong conclusion, then turn it back on before the workflow writes anywhere a human reads, such as a Slack message.
03

Maximum Number of Tokens

options.maxTokens

What you see in n8n

Notes & use cases

Caps the length of the generated completion. A token is the unit models count text in, roughly a word fragment, so this is a budget on the answer, not on your prompt.

Key parameters

  • Maximum Number of Tokens: a number. The ceiling you can ask for depends on the model chosen above, so the same value is not accepted everywhere in the list.
Use cases
keep it tight when the answer feeds a field with a display limit, and raise it for a Summarization Chain that has to return several paragraphs instead of a line.
04

Sampling Temperature

options.temperature

What you see in n8n

Notes & use cases

Controls how random the sampling is. Push it up and the wording varies between runs; bring it toward zero and the model turns deterministic and repetitive.

Key parameters

  • Sampling Temperature: a number. A higher value gives more diverse output and raises the risk of hallucinations.
Use cases
near zero for anything a later node parses, like a label or a status field written back to a sheet. Higher only when the text is read by a person and repetition would be noticeable, such as a first draft in Gmail.
05

Timeout

options.timeout

What you see in n8n

Notes & use cases

Sets how long a single request is allowed to take, in milliseconds, before the node stops waiting. Long generations and slow moments on the provider side are what this guards against.

Key parameters

  • Timeout: a number in milliseconds, so 2000 means two seconds of patience.
Use cases
a short value in a workflow that answers a person live, where a late reply is worthless anyway. A generous one for a nightly batch, where finishing matters more than finishing fast.
06

Max Retries

options.maxRetries

What you see in n8n

Notes & use cases

Says how many times the node tries the request again after a failure. Beyond that count the execution fails and the root node fails with it.

Key parameters

  • Max Retries: a number of attempts, counted after the first call.
Use cases
raise it on a loop that processes hundreds of rows, where one refused call would otherwise stop the whole run. Keep it low on an interactive path: retrying twice behind a Timeout that is already generous makes the wait feel much longer than the numbers suggest.
07

Top P

options.topP

What you see in n8n

Notes & use cases

Restricts the pool the model samples from, by probability rather than by randomness. At 0.5, half of all likelihood-weighted options are considered and the rest are dropped.

Key parameters

  • Top P: a number. Lower it to ignore the less probable continuations.
Use cases
the usual advice is to adjust this or Sampling Temperature, not both at once, so pick the one you will document for the team and leave the other alone.
Need help

Need help automating MiniMax Chat Model with n8n?

A person reads every message.

FAQ

Questions people ask next

01Is the n8n MiniMax Chat Model node free to use?
Yes, on the n8n side. The node ships inside the AI package that comes with n8n, so there is nothing to install and no extra charge, on n8n Cloud as on a self-hosted instance running the Community Edition under the Sustainable Use license. What is not free is the model itself: the provider bills its own API calls according to its own terms, and n8n adds nothing on top of that. On n8n Cloud you can also run the node on Gateway credits instead of your own key, which moves where the usage is counted.
02What credentials do you need to run it?
A MiniMax credential, created once from the Credentials menu and reusable across every workflow on the instance. The node itself has no authentication selector: you attach the credential to the node and that is the whole setup. On n8n Cloud there is an alternative, Use Gateway credits in the node's credential field, which lets the node run without a MiniMax account. Note that the credential belongs to this sub-node, not to the agent or chain it feeds, which is where most setup confusion comes from.
03What are the limits of this node?
It only does one thing: hand a MiniMax completion to the root node above it. It carries no memory and no tools, because in n8n those are separate sub-nodes on separate ports. The node is at version 1, with 7 parameters and a single output port, so anything outside that surface is out of reach here. The model dropdown is fixed to the entries shipped with the node, and the maximum number of tokens you can request depends on which of those you selected. An older workflow may still show an earlier version with fewer options.
04What do you need to plug in at minimum to get an answer?
A root node and this one, wired together. Drop an AI Agent or a Basic LLM Chain on the canvas, attach the MiniMax Chat Model node to its Model port, set the credential and pick an entry in Model. That pair already runs. Start with Basic LLM Chain rather than AI Agent if the job is a single prompt: there is less to configure and less to misread when something fails. Memory and tools are extra sub-nodes you add later, on the root node's other ports, once the plain prompt behaves.
05n8n or Make for an AI workflow like this?
It depends on where you want the data to sit and how you want to pay. Make is hosted only and billed per operation, which is predictable and needs no server from you. n8n runs the same workflow on Cloud or on your own infrastructure through Docker or npm, so an AI workflow handling customer messages can stay inside your own environment. The visual logic differs too: n8n's root node plus sub-node model is explicit about which model, memory and tools an agent uses, which is easier to audit and a little more to wire.
Hack'celeration Lab

Get our weekly integration tips.

No spam. Unsubscribe anytime.

Same platform

Other docs you might need on this stack.

Browse all integrations