GPT-5 Mini

The compact model that still speaks JSON

By Runware·Created 2 years ago·Updated about a month ago· 140

GPT-5 Mini

messages
images

text

◄seed0►

◄numberResults1►

◄includeUsagefalse►

◄settings.maxTokens32000►

◄settings.systemPrompt►

◄settings.thinkingLevelmedium►

◄toolChoicefalse►

◄toolChoice.name►

◄toolChoice.type(default)►

◄outputFormatTEXT►

◄advanced_json►

Between the 5.4 family and the 5.5 flagship sits GPT-5 Mini, and it occupies the sweet spot a lot of workflows actually want: a smaller model that's cheaper per call than 5.5, thinks by default but lets you turn it off, and - same as 5.5 - offers a JSON output mode for feeding structured data back into your graph. If you want "5.5-ish behavior without the 5.5 price," this is the one to reach for.

How it works

A textInference node calling GPT-5 Mini through the Runware cloud. Required messages input is a RUNWARE_MESSAGES socket from the Runware Messages builder (Runware/Params).

The layout is the reasoning-model style, with a few twists vs. the 5.4 nodes:

settings.thinkingLevel - defaults to medium, with none/low/medium/high (no xhigh - the Mini doesn't get the deepest reasoning tier). Turning thinking off for trivial tasks is where the savings live.
settings.maxTokens - default 32000 (up to 128k), noticeably more headroom than the 5.4 nodes' 4096 default, so Mini is comfortable with long outputs.
outputFormat - TEXT or JSON. JSON mode plus advanced_json (which accepts jsonSchema and tools definitions) makes this a workable structured-output node for a fraction of the flagship's cost.
settings.systemPrompt, seed, images (IMAGE) for vision, toolChoice / toolChoice.type / toolChoice.name, numberResults, and includeUsage.

Note there's no temperature/topP here - same policy as the other reasoning-tier models. Output is a text STRING (parseable JSON when JSON mode is on).

Install

Part of ComfyUI-Runware:

cd ComfyUI/custom_nodes
git clone https://github.com/Runware/ComfyUI-Runware
pip install -r ComfyUI-Runware/requirements.txt

Restart (or ComfyUI Manager → search "Runware"). Cloud model, nothing to download. API key: Settings → Runware API key, RUNWARE_API_KEY, or runware auth login.

Gotchas

"Mini" here means smaller, not small - with thinking on and a big maxTokens it can still run a bill up, so check the title-bar cost. The missing xhigh thinking tier is the real difference from 5.5; if you need the deepest chains, that's a 5.5/Pro job. And as with every text node in this pack, executions are stateless - no conversation memory, so build multi-turn by feeding the text output back into the Messages builder yourself.

CategoryRunware/Text/openai

Inputs (13)

Name	Type	Default	Description
messages	RUNWARE_MESSAGES		—
imagesopt	IMAGE		—
seedopt	INT	00–2147483647	Random seed for reproducible generation. When not provided, a random seed is generated in the unsigned 32-bit range.
numberResultsopt	INT	11–4	Number of results to generate. Each result uses a different seed, producing variations of the same parameters.
includeUsageopt	BOOLEAN	false	Include token usage statistics in the response.
settings.maxTokensopt	INT	320001–128000	Maximum number of tokens to generate in the response.
settings.systemPromptopt	STRING		System-level instruction that guides the model's behavior and output style across the entire generation.
settings.thinkingLevelopt	COMBO	medium	Controls the depth of internal reasoning the model performs before generating a response.
toolChoiceopt	BOOLEAN	false	Enable to set toolChoice. Off uses the model's default.
toolChoice.nameopt	STRING		Name of the specific tool the model must call. Required when type is `tool`.
toolChoice.typeopt	COMBO	(default)	Strategy the model uses to decide when and which tools to call.
outputFormatopt	COMBO	TEXT	Output format for the generated text.
advanced_jsonopt	STRING		Optional JSON merged into the request. For: jsonSchema, tools

Outputs (1)

Name	Type	Description
text	STRING	—