[Integrations](https://spiral.ai/integrations) › [Hermes](https://spiral.ai/integrations/hermes) › [GPT-5.4 mini](https://spiral.ai/integrations/hermes/gpt-5-4-mini)

# How to Use GPT-5.4 mini in Hermes

Use GPT-5.4 mini for targeted code edits, debugging, and screenshot interpretation. It is a faster alternative to GPT-5.4 when you can give the task a clear scope and explicit instructions.

|  |  |
| --- | --- |
| Model | GPT-5.4 mini |
| Available through | OpenAI |
| Category | LLM |
| Price | Up to $0.023997 per run at the default inputs. $0.975 per 1M input tokens, $0.0975 per 1M cached input tokens, $5.85 per 1M output tokens. |
| Returns | text, json |
| Runs | In the background |
| Setup | About 2 minutes, no API key |

## Prompts to try with GPT-5.4 mini in Hermes

- On each daily run, use GPT-5.4 mini to review newly received bug reports. For each report, extract observed behavior, expected behavior, and reproduction steps. Flag missing details and return a concise review queue.
- Each morning, use GPT-5.4 mini to inspect the latest UI captures supplied to this task against the reference captures. Report visible changes by screen and identify unreadable areas. If a capture is missing, report the gap.
- On each run, use GPT-5.4 mini to review newly supplied code changes for empty-input handling. Return only findings supported by the change, with the affected function and a concrete failure example. Report no findings when none are supported.
- Every weekday, use GPT-5.4 mini to turn the latest supplied debugging notes into a handoff. Include the reproduced failure, attempted fixes, observed results, and next unresolved check. Keep the report under 300 words.
- On each scheduled run, use GPT-5.4 mini with high reasoning effort to assess the supplied failing-test results and proposed patches. Check whether each patch addresses its stated failure. Return supported conclusions and unresolved questions for review.

## How to set up GPT-5.4 mini in Hermes

Pick the way that suits you. Each takes about two minutes.

### Use the Spiral skill

1. **Send Hermes this message**: Your agent reads Spiral's skill and sets itself up.

```
set up https://spiral.ai/SKILL.md
```

2. **Sign in or paste a key**: Approve Spiral in the browser, or give the agent a key from the console.

### Use Spiral's MCP server

1. **Add the server and sign in**: Hermes opens your browser: sign in to Spiral and click Authorize. No API key. When Hermes asks Enable all tools? \[Y/n/select], answer Y: n cancels without saving the server, and select hides the tools you leave out.

```bash
hermes mcp add spiral \
  --url https://api.spiral.ai/mcp \
  --auth oauth \
  --connect-timeout 300
```

2. **Check it**: In a chat that was already open, run `/reload-mcp`.

```bash
hermes mcp list
hermes mcp test spiral
```

### Use the Spiral CLI

1. **Install and sign in**: Installs the Spiral CLI, signs you in and adds its skill to your coding agents.

```bash
npx -y spiral-cli@latest init
```

### Call the Spiral API

1. **Create an API key**: In the console, under API keys.

2. **Run GPT-5.4 mini**: One request; the result comes back by polling or a webhook.

```bash
curl https://api.spiral.ai/v1/runs \
  -H "Authorization: Bearer $SPIRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "provider": "openai",
    "model": "openai/gpt-5.4-mini",
    "input": {
      "messages": [
        {
          "role": "system"
        }
      ]
    }
  }'
```

Then paste one of the prompts above: Hermes finds GPT-5.4 mini on Spiral, shows the price and runs it.

Setup guide in the docs: https://spiral.ai/docs/quickstarts/hermes

## GPT-5.4 mini tools and prices

- [GPT-5.4 Mini](https://spiral.ai/tools/openai/openai/gpt-5.4-mini) (OpenAI): Up to $0.023997 per run at the default inputs. $0.975 per 1M input tokens, $0.0975 per 1M cached input tokens, $5.85 per 1M output tokens.

## Settings you can ask for

| Setting | Allowed values | Default | Description |
| --- | --- | --- | --- |
| seed |  | — |  |
| tools |  | — |  |
| top_p | … to 1 | — | Taken only at reasoning effort "none", this model's default. |
| logprobs |  | — | Taken only at reasoning effort "none", this model's default. |
| reasoning |  | — | Reasoning effort; reasoning tokens count toward max_tokens. |
| max_tokens | 16 to 128000 | — | The most output tokens to generate. |

## What can you do with GPT-5.4 mini in Hermes?

GPT-5.4 mini can apply the same focused review to incoming bug reports, code changes, or UI captures on each scheduled run. Define the evidence it should inspect, the report format, and how it should handle missing information.

### Fix a reproducible bug

Give GPT-5.4 mini the failing behavior, relevant code, and expected result. Ask for a focused patch and a check that reproduces the original failure.

### Find where a feature lives

Ask GPT-5.4 mini to trace a feature through a codebase and return the relevant files, functions, and call sequence.

### Build a front-end component

Specify the component's states, interactions, and existing styling conventions. Have GPT-5.4 mini implement the component and check each required behavior.

### Explain a confusing screen

Provide an upright screenshot and ask GPT-5.4 mini to identify visible controls and explain a specific task. Require it to flag unreadable labels.

### Turn bug reports into structured records

Have GPT-5.4 mini extract reproduction steps, observed behavior, and expected behavior into a defined JSON structure. Preserve missing details as unknown.

### Review a bounded code change

Assign GPT-5.4 mini one review question, such as whether an edit handles empty input. Request findings with code references and a concrete failure example.

### Prepare a UI regression report

Give GPT-5.4 mini before-and-after screenshots and a list of expected changes. Ask it to report visible differences and separate uncertain observations.

## Overview

- [OpenAI](https://spiral.ai/providers/openai)
- [All Hermes guides](https://spiral.ai/integrations/hermes)

## GPT-5.4 mini in other agents

- [Amp](https://spiral.ai/integrations/amp/gpt-5-4-mini)
- [ChatGPT](https://spiral.ai/integrations/chatgpt/gpt-5-4-mini)
- [Claude](https://spiral.ai/integrations/claude-desktop/gpt-5-4-mini)
- [Claude Code](https://spiral.ai/integrations/claude-code/gpt-5-4-mini)
- [Cline](https://spiral.ai/integrations/cline/gpt-5-4-mini)
- [Codex](https://spiral.ai/integrations/codex/gpt-5-4-mini)
- [Cursor](https://spiral.ai/integrations/cursor/gpt-5-4-mini)
- [Devin](https://spiral.ai/integrations/devin/gpt-5-4-mini)
- [dots](https://spiral.ai/integrations/dots/gpt-5-4-mini)
- [Factory Droid](https://spiral.ai/integrations/factory-droid/gpt-5-4-mini)
- [Gemini](https://spiral.ai/integrations/gemini/gpt-5-4-mini)
- [Gemini CLI](https://spiral.ai/integrations/gemini-cli/gpt-5-4-mini)
- [Google Antigravity](https://spiral.ai/integrations/antigravity/gpt-5-4-mini)
- [goose](https://spiral.ai/integrations/goose/gpt-5-4-mini)
- [Grok](https://spiral.ai/integrations/grok/gpt-5-4-mini)
- [Grok Build](https://spiral.ai/integrations/grok-build/gpt-5-4-mini)
- [Instinct](https://spiral.ai/integrations/instinct/gpt-5-4-mini)
- [Manus](https://spiral.ai/integrations/manus/gpt-5-4-mini)
- [Muse](https://spiral.ai/integrations/muse/gpt-5-4-mini)
- [Muse Code](https://spiral.ai/integrations/muse-code/gpt-5-4-mini)
- [n8n](https://spiral.ai/integrations/n8n/gpt-5-4-mini)
- [Notion](https://spiral.ai/integrations/notion/gpt-5-4-mini)
- [OpenClaw](https://spiral.ai/integrations/openclaw/gpt-5-4-mini)
- [opencode](https://spiral.ai/integrations/opencode/gpt-5-4-mini)
- [Perplexity](https://spiral.ai/integrations/perplexity/gpt-5-4-mini)
- [Raycast](https://spiral.ai/integrations/raycast/gpt-5-4-mini)
- [VS Code](https://spiral.ai/integrations/vscode/gpt-5-4-mini)
- [Windsurf](https://spiral.ai/integrations/windsurf/gpt-5-4-mini)
- [Zapier](https://spiral.ai/integrations/zapier/gpt-5-4-mini)

## More Hermes guides

- [GPT-5.4 Pro](https://spiral.ai/integrations/hermes/gpt-5-4-pro)
- [GPT-5.5](https://spiral.ai/integrations/hermes/gpt-5-5)
- [GPT-5.4 Nano](https://spiral.ai/integrations/hermes/gpt-5-4-nano)
- [GPT-5.5 Pro](https://spiral.ai/integrations/hermes/gpt-5-5-pro)
- [GPT-5.4](https://spiral.ai/integrations/hermes/gpt-5-4)
- [GPT-5.6 Luna](https://spiral.ai/integrations/hermes/gpt-5-6-luna)

## Questions about GPT-5.4 mini in Hermes

**Can Hermes use GPT-5.4 mini?**

Yes. Add Spiral to Hermes as an MCP server, and Hermes can find and run GPT-5.4 mini like any built-in tool.

**How much does GPT-5.4 mini cost in Hermes?**

Up to $0.023997 per run at the default inputs. $0.975 per 1M input tokens, $0.0975 per 1M cached input tokens, $5.85 per 1M output tokens. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Hermes plan is separate.

**Do I need an OpenAI account to use GPT-5.4 mini in Hermes?**

No. Spiral runs GPT-5.4 mini on its own OpenAI account. You sign in to Spiral once and never paste a key.

**How long does it take to set up GPT-5.4 mini in Hermes?**

About two minutes: add the server and sign in, then check it.

**Does Hermes wait while GPT-5.4 mini runs?**

GPT-5.4 mini runs in the background. Hermes starts the run, checks on it and gives you the text, json when it's ready.

**What is GPT-5.4 mini good at?**

Targeted code changes, finding relevant code, generating front-end code, and debugging are strong starting points. It also handles screenshot interpretation for computer use tasks.

**How does GPT-5.4 mini differ from GPT-5.4?**

GPT-5.4 mini offers faster responses in a smaller model. Give it a clearly structured task, since OpenAI identifies implicit workflows and ambiguity handling as weaker areas.

**How much text can GPT-5.4 mini handle?**

Its context window is 400,000 tokens, and its maximum output is 128,000 tokens. Those are token limits, so avoid treating them as fixed word or page counts.

**How should I prompt GPT-5.4 mini?**

State the task, the required steps, and the expected answer format explicitly. Tell it what to do when information is missing, such as asking a question or returning an unknown value.

**When should I increase reasoning effort?**

Try higher effort for debugging or decisions that require several dependent checks. Compare the results on representative tasks before adopting the change; extra reasoning is a setting to test, not a guarantee of correctness.

