RecipeAgents / n8n nodes / OpenAI Chat Model
Node reference · data pulled Oct 26, 2026n8n OpenAI Chat Model node: what the data shows
Plugs an OpenAI model into an agent or chain as its brain.
What it does
This is a sub-node: it hangs under an AI Agent, LLM chain or extractor and supplies the model. It is not the same as the standalone OpenAI node, which calls the API directly as a step. The model list is loaded from your account, and it can use OpenAI's newer Responses API with built-in tools; on n8n Cloud it can also run on n8n's own credits instead of your key.
Most templates using it (about four in five) attach it to an AI Agent. Our own recipes mostly use the standalone OpenAI node instead, because they make one call and don't need an agent.
How often public templates use it
3,104 of the 13,574 public templates on n8n.io (22.9%) contain this node, in the snapshot we took on 2026-10-20. The table splits that by the year each template was published. 2026 runs to October 20.
| Year published | Templates with the node | Share of that year |
|---|---|---|
| 2024 | 94 of 462 | 20.3% |
| 2025 | 1,774 of 6,564 | 27.0% |
| 2026 | 1,220 of 5,845 | 20.9% |
The median template using this node has 419 views on n8n.io, against 284 for all templates. Template counts show what people publish, not what runs in production.
What it is used with
The most frequent partners are the nodes that appear most often in the same templates. "Lift" compares that with the whole library: 2× means the partner turns up twice as often next to this node as in an average template.
| Most frequent partner | Share of its templates | Lift |
|---|---|---|
| AI Agent | 79.2% | 2.53× |
| Edit Fields (Set) | 63.7% | 1.14× |
| Code | 54.2% | 0.91× |
| If | 48.6% | 0.96× |
| HTTP Request | 47.0% | 0.91× |
| Structured Output Parser | 39.0% | 2.43× |
Strong pairings (in at least 5% of its templates, lift 1.5× or more): 13. The top ones:
| Strong partner | Share of its templates | Lift |
|---|---|---|
| Embeddings OpenAI | 8.4% | 3.27× |
| Call n8n Workflow Tool | 5.4% | 2.80× |
| Information Extractor | 6.1% | 2.76× |
| Simple Memory | 21.8% | 2.73× |
| AI Agent | 79.2% | 2.53× |
| Default Data Loader | 7.2% | 2.44× |
| Structured Output Parser | 39.0% | 2.43× |
| Chat Trigger | 13.5% | 2.41× |
Most viewed public templates that use it: Ai agent chat (1,847,731 views); Building your first whatsapp chatbot (541,254 views); Automate multi platform social media content creation with ai (442,651 views).
Versions and changes
Node versions from the npm packages n8n-nodes-base and @n8n/n8n-nodes-langchain, version 2.44.4. New workflows get the default version; workflows built earlier keep the version they were built with until you replace the node.
| Node | Versions | Default | Credential types | Scope |
|---|---|---|---|---|
| OpenAI Chat Model | 1, 1.1, 1.2, 1.3 | 1.3 | OpenAI API key | – |
Source on GitHub: packages/@n8n/nodes-langchain/nodes/llms/LMChatOpenAi: 63 commits since 2023-11-29, last on 2026-10-05.
Recent fixes and features
- 2026-10-05: fix(Azure OpenAI Chat Model Node): Support Responses-only deployments (PR #39403)
- 2026-07-29: feat(OpenAI Chat Model Node): Add optional extraBody option (PR #13992)
- 2025-06-05: feat(OpenAI Chat Model Node): Update default model to gpt-4.1-mini (PR #15975)
- 2025-03-13: fix(OpenAI Chat Model Node): Sort models alphabetically (PR #13909)
- 2025-02-06: feat(OpenAI Chat Model Node): Add reasoning effort option to control the amount of reasoning tokens to use (PR #13103)
- 2025-02-03: feat(OpenAI Chat Model Node, OpenAI Node): Include o3 models in model selection (PR #13005)
Merged dates from GitHub. A fix only helps once your n8n instance runs a release that includes it, so check your version.
Known problems
- Expressions only read the first item. Sub-node parameters resolve against the first incoming item, not each item, so per-item settings don't work as expected. Source: n8n docs
- Default model changed. In June 2025 the default model became gpt-4.1-mini; templates that relied on the default may now use a different model than their author did. Source: n8n PR #15975
- Reasoning effort. A reasoning-effort option was added in 2025 for reasoning models; leaving it unset uses the provider default. Source: n8n PR #13103
Forum trouble signal: not measured. The n8n community forum's terms forbid automated access, so we don't count forum threads.
Our recipes that use it
1 of our 25 importable workflows uses this node:
Official docs and sources
Usage, pairings, AI share and views: our count of 13,574 public n8n.io templates (snapshot 2026-10-20). Versions: npm packages, version 2.44.4. Changes: n8n's GitHub repository. Docs read and data pulled Oct 26, 2026. We describe each node in our own words; the docs are the authority on settings.
Related nodes
- Simple Memory1,086 templates · score 71.7
- AI Agent4,246 templates · score 69.9
- Structured Output Parser2,181 templates · score 68.3
- Google Gemini Chat Model1,285 templates · score 65.1
- All 22 nodesSortable table with the scoring method
Questions
- How many n8n templates use the OpenAI Chat Model node?
- 3,104 of the 13,574 public templates on n8n.io (22.9%), in our snapshot of 2026-10-20.
- What is the OpenAI Chat Model node most often used with?
- AI Agent, which appears in 79.2% of its templates. The most distinctive partner is Embeddings OpenAI (lift 3.3×).
- OpenAI Chat Model or OpenAI node?
- The Chat Model is a sub-node for agents and chains. The OpenAI node is a normal step that sends one request. Use the second for single prompts.
- Can I use OpenAI-compatible providers?
- The credential takes a base URL, so compatible endpoints can work, but model lists and features vary.