[Integrations](https://spiral.ai/integrations) › [Amp](https://spiral.ai/integrations/amp) › [Gemini 4 Neon](https://spiral.ai/integrations/amp/gemini-4-neon)

# How to Use Gemini 4 Neon in Amp

Gemini 4 Neon suits multi-file debugging, architectural refactoring, and analytical work with several steps. It sits between Gemini 4 Carbon for the deepest long projects and Gemini 4 Hydrogen for fast everyday execution.

|  |  |
| --- | --- |
| Model | Gemini 4 Neon |
| Available through | Vertex AI |
| Category | LLM |
| Price | Up to $0.080009 per run at the default inputs. $3.90 per 1M input tokens, $0.195 per 1M cached input tokens, $0.108333335 per 1M cache write tokens, $19.50 per 1M output tokens. |
| Returns | text, json |
| Runs | In the background |
| Setup | About 2 minutes, no API key |

## Prompts to try with Gemini 4 Neon in Amp

- Use Gemini 4 Neon with high reasoning to investigate the failing tests in this repository. Trace the root cause, make the smallest justified fix, and run the relevant tests. Report what passed and what remains unverified.
- Use Gemini 4 Neon with high reasoning to map this project's authentication flow. Identify the files involved, explain where authorization decisions happen, and propose a staged refactor before editing.
- Use Gemini 4 Neon to review the current diff for correctness. For each suspected bug, give the file location, triggering input, expected behavior, and a focused test that would expose it.
- Use Gemini 4 Neon with high reasoning to remove duplicated validation logic in this project. Preserve existing behavior, update callers, and run tests covering invalid and boundary inputs.
- Use Gemini 4 Neon with low reasoning to summarize this repository's recent commits into release notes. Group changes by user impact and include commit references for each item.

## How to set up Gemini 4 Neon in Amp

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

### Use Spiral's MCP server

1. **Add the server**: No header and no client ID or secret: Amp registers itself with Spiral.

```bash
amp mcp add spiral https://api.spiral.ai/mcp
```

2. **Sign in**: Start `amp`. Amp opens Spiral's sign-in page in your browser: sign in, pick the organization and click Authorize. No API key.

3. **Check it**

```bash
amp mcp doctor
```

### Use the Spiral skill

1. **Send Amp 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 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 Gemini 4 Neon**: 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": "vertex",
    "model": "google/gemini-4-neon",
    "input": {
      "messages": [
        {
          "role": "user",
          "content": "Hello!"
        }
      ]
    }
  }'
```

Then paste one of the prompts above: Amp finds Gemini 4 Neon on Spiral, shows the price and runs it.

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

## Gemini 4 Neon tools and prices

- [Gemini 4 Neon](https://spiral.ai/tools/vertex/google/gemini-4-neon) (Vertex AI): Up to $0.080009 per run at the default inputs. $3.90 per 1M input tokens, $0.195 per 1M cached input tokens, $19.50 per 1M output tokens.

## Settings you can ask for

| Setting | Allowed values | Default | Description |
| --- | --- | --- | --- |
| seed |  | — |  |
| stop |  | — |  |
| tools |  | — |  |
| reasoning |  | — | Reasoning effort; reasoning tokens count toward max_tokens. |
| max_tokens | 1 to 65536 | — | The most output tokens to generate. On a reasoning model the reasoning tokens count toward it: a cap too low for the effort can end with no answer (finish_reason length). |
| tool_choice |  | — |  |

## What can you do with Gemini 4 Neon in Amp?

Turn repository context into targeted fixes and staged refactors with Gemini 4 Neon. Ask the agent to trace dependencies, make bounded changes, and verify the result with relevant tests.

### Trace a bug across files

Follow a failing request through its callers, identify the root cause, and propose a small patch with a regression test.

### Plan an architectural refactor

Map dependencies before changing a shared interface. Request a staged migration plan with compatibility checks and rollback points.

### Review a proposed code change

Inspect a diff for behavior changes and missing edge cases. Ask for a reproducible example for each suspected defect.

### Compare technical proposals

Evaluate supplied proposals against explicit requirements. Separate documented support, assumptions, and unanswered questions.

### Summarize operational incidents

Turn supplied event records into a timeline, impact summary, and follow-up list. Preserve uncertainty about causes.

### Route incoming requests

Classify request text using a fixed category list and escalation rules. Include the supporting passage for each assignment.

### Extract action items

Convert meeting text into tasks with owners and deadlines. Mark missing details explicitly instead of filling them in.

### Build an evidence-based decision brief

Compare options using supplied research and constraints. Request a recommendation, its supporting evidence, and conditions that would change it.

## Overview

- [Google Vertex AI](https://spiral.ai/providers/vertex)
- [All Amp guides](https://spiral.ai/integrations/amp)

## Gemini 4 Neon in other agents

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

## More Amp guides

- [Gemini 4 Hydrogen Vertex AI](https://spiral.ai/integrations/amp/gemini-4-hydrogen-vertex-ai)
- [Gemini Flash Latest](https://spiral.ai/integrations/amp/gemini-flash-latest)
- [Gemini 4 Hydrogen](https://spiral.ai/integrations/amp/gemini-4-hydrogen)
- [Gemini Pro Latest](https://spiral.ai/integrations/amp/gemini-pro-latest)
- [Gemini 4 Carbon Batch](https://spiral.ai/integrations/amp/gemini-4-carbon-batch)
- [Gemma 2 27B Instruct](https://spiral.ai/integrations/amp/gemma-2-27b-instruct)

## Questions about Gemini 4 Neon in Amp

**Can Amp use Gemini 4 Neon?**

Yes. Add Spiral to Amp as an MCP server, and Amp can find and run Gemini 4 Neon like any built-in tool.

**How much does Gemini 4 Neon cost in Amp?**

Up to $0.080009 per run at the default inputs. $3.90 per 1M input tokens, $0.195 per 1M cached input tokens, $0.108333335 per 1M cache write tokens, $19.50 per 1M output tokens. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Amp plan is separate.

**Do I need a Vertex AI account to use Gemini 4 Neon in Amp?**

No. Spiral runs Gemini 4 Neon on its own Vertex AI account. You sign in to Spiral once and never paste a key.

**How long does it take to set up Gemini 4 Neon in Amp?**

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

**Does Amp wait while Gemini 4 Neon runs?**

Gemini 4 Neon runs in the background. Amp starts the run, checks on it and gives you the text, json when it's ready.

**What is Gemini 4 Neon good at?**

Gemini 4 Neon is built for complex coding and analytical work. Practical tasks include tracing bugs across files, planning architectural changes, and turning supplied evidence into a decision brief.

**How does Gemini 4 Neon compare with Carbon and Hydrogen?**

Google positions Gemini 4 Carbon for the deepest long projects and Gemini 4 Hydrogen for fast everyday execution. Gemini 4 Neon sits between them for demanding coding and analytical tasks.

**How much context can Gemini 4 Neon handle?**

The total budget is 1,048,576 tokens shared between input and output. Leave room for the response when supplying a large amount of material.

**How long can a Gemini 4 Neon answer be?**

Google documents a 65,536-token limit for a single foreground segment. That budget includes thinking and function calls as well as the visible answer, so it is not a promise of 65,536 tokens of prose.

**When should I use more reasoning?**

Use high reasoning for architectural refactoring, code generation, or analysis with several dependent steps. Low reasoning suits classification, summaries, and extracting structured details.

