[Integrations](https://spiral.ai/integrations) › [Claude](https://spiral.ai/integrations/claude-desktop) › [MiniMax M3 (batch)](https://spiral.ai/integrations/claude-desktop/minimax-m3-batch)

# How to Use MiniMax M3 (batch) in Claude

MiniMax M3 suits repository analysis, bug fixes, and research across long documents. It combines a million-token context window with tool use and optional reasoning for tasks that require several steps.

|  |  |
| --- | --- |
| Model | MiniMax M3 (batch) |
| Available through | OpenRouter |
| Category | LLM |
| Price | Up to $0.004697 per run at the default inputs. $0.2912 per 1M input tokens, $1.144 per 1M output tokens. |
| Returns | text, json |
| Runs | In the background |
| Setup | About 2 minutes, no API key |

## Prompts to try with MiniMax M3 (batch) in Claude

- Use MiniMax M3 to compare the specifications pasted above. Return a requirements matrix, cite the relevant sections, and flag contradictions without resolving them by guesswork.
- Use MiniMax M3 to analyze this error report and pasted code. Explain the likely cause, propose a minimal correction, and give me a reproduction check.
- Use MiniMax M3 to research the technical question discussed above using available search tools. Prioritize original documentation and return a 600-word brief with sources and open questions.
- Use MiniMax M3 to turn this incident timeline into a report. Separate observed events from hypotheses, list unresolved questions, and suggest checks for each hypothesis.
- Use MiniMax M3 to compare the two architecture proposals above. Evaluate them against the stated requirements and finish with a recommendation that names its assumptions.

## How to set up MiniMax M3 (batch) in Claude

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

### Use the Claude connector

1. **Add Spiral to Claude**: Click Add Spiral in Claude above. Claude opens Add custom connector with Spiral and its URL filled in; check them and click Continue. Without the link: in Claude open Customize › Connectors › Add › Add custom connector and paste this URL:

```text
https://api.spiral.ai/mcp
```

2. **Keep the detected sign-in and click Add**: Claude finds Spiral's sign-in by itself: leave Sign in now and Use Claude's published identity selected.

3. **Click Connect**: The connector page opens. Connect takes you to Spiral in your browser.

### Use the Spiral skill

1. **Send Claude 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 MiniMax M3 (batch)**: 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": "openrouter",
    "model": "minimax/minimax-m3:batch",
    "input": {
      "messages": [
        {
          "role": "system"
        }
      ]
    }
  }'
```

Then paste one of the prompts above: Claude finds MiniMax M3 (batch) on Spiral, shows the price and runs it.

Setup guide in the docs: https://spiral.ai/docs/quickstarts/claude-desktop

## MiniMax M3 (batch) tools and prices

- [MiniMax: MiniMax M3 (batch)](https://spiral.ai/tools/openrouter/minimax/minimax-m3%3Abatch) (OpenRouter): Up to $0.004697 per run at the default inputs. $0.2912 per 1M input tokens, $1.144 per 1M output tokens.

## Settings you can ask for

| Setting | Allowed values | Default | Description |
| --- | --- | --- | --- |
| seed |  | — |  |
| stop |  | — |  |
| min_p | 0 to 1 | — |  |
| tools |  | — |  |
| top_k | 0 to … | — |  |
| top_p | 0 to 1 | — |  |

## What can you do with MiniMax M3 (batch) in Claude?

Turn long technical discussions and source material into decisions you can check with MiniMax M3. Ask for evidence references, explicit assumptions, and a deliverable sized for the decision at hand.

### Trace a bug across a repository

Follow an error from its entry point through related modules, propose a focused patch, and check it against a reproduction.

### Plan a code migration

Compare existing code with a target specification, identify affected components, and produce an ordered migration plan with acceptance checks.

### Investigate a performance bottleneck

Review code and profiling evidence, propose a measurable change, and compare correctness and timing before accepting it.

### Compare long technical documents

Build a requirements matrix across specifications, flag contradictions, and cite the sections behind each finding.

### Prepare a research brief

Use search results and source material to answer a narrow question, explain disagreements, and list evidence that remains missing.

### Turn operational records into a report

Combine supplied status updates, incident timelines, and action lists into a concise report with owners and unresolved issues.

## Overview

- [OpenRouter](https://spiral.ai/providers/openrouter)
- [All Claude guides](https://spiral.ai/integrations/claude-desktop)

## MiniMax M3 (batch) in other agents

- [Amp](https://spiral.ai/integrations/amp/minimax-m3-batch)
- [ChatGPT](https://spiral.ai/integrations/chatgpt/minimax-m3-batch)
- [Claude Code](https://spiral.ai/integrations/claude-code/minimax-m3-batch)
- [Cline](https://spiral.ai/integrations/cline/minimax-m3-batch)
- [Codex](https://spiral.ai/integrations/codex/minimax-m3-batch)
- [Cursor](https://spiral.ai/integrations/cursor/minimax-m3-batch)
- [Devin](https://spiral.ai/integrations/devin/minimax-m3-batch)
- [dots](https://spiral.ai/integrations/dots/minimax-m3-batch)
- [Factory Droid](https://spiral.ai/integrations/factory-droid/minimax-m3-batch)
- [Gemini](https://spiral.ai/integrations/gemini/minimax-m3-batch)
- [Gemini CLI](https://spiral.ai/integrations/gemini-cli/minimax-m3-batch)
- [Google Antigravity](https://spiral.ai/integrations/antigravity/minimax-m3-batch)
- [goose](https://spiral.ai/integrations/goose/minimax-m3-batch)
- [Grok](https://spiral.ai/integrations/grok/minimax-m3-batch)
- [Grok Build](https://spiral.ai/integrations/grok-build/minimax-m3-batch)
- [Hermes](https://spiral.ai/integrations/hermes/minimax-m3-batch)
- [Instinct](https://spiral.ai/integrations/instinct/minimax-m3-batch)
- [Manus](https://spiral.ai/integrations/manus/minimax-m3-batch)
- [Muse](https://spiral.ai/integrations/muse/minimax-m3-batch)
- [Muse Code](https://spiral.ai/integrations/muse-code/minimax-m3-batch)
- [n8n](https://spiral.ai/integrations/n8n/minimax-m3-batch)
- [Notion](https://spiral.ai/integrations/notion/minimax-m3-batch)
- [OpenClaw](https://spiral.ai/integrations/openclaw/minimax-m3-batch)
- [opencode](https://spiral.ai/integrations/opencode/minimax-m3-batch)
- [Perplexity](https://spiral.ai/integrations/perplexity/minimax-m3-batch)
- [Raycast](https://spiral.ai/integrations/raycast/minimax-m3-batch)
- [VS Code](https://spiral.ai/integrations/vscode/minimax-m3-batch)
- [Windsurf](https://spiral.ai/integrations/windsurf/minimax-m3-batch)
- [Zapier](https://spiral.ai/integrations/zapier/minimax-m3-batch)

## More Claude guides

- [MiniMax M3](https://spiral.ai/integrations/claude-desktop/minimax-m3)
- [MiniMax M3.1 Flash](https://spiral.ai/integrations/claude-desktop/minimax-m3-1-flash)
- [MiniMax M2.7](https://spiral.ai/integrations/claude-desktop/minimax-m2-7)
- [Ministral 3 14B](https://spiral.ai/integrations/claude-desktop/ministral-3-14b)
- [MiniMax M2.5](https://spiral.ai/integrations/claude-desktop/minimax-m2-5)
- [Ministral 3 3B](https://spiral.ai/integrations/claude-desktop/ministral-3-3b)

## Questions about MiniMax M3 (batch) in Claude

**Can Claude use MiniMax M3 (batch)?**

Yes. Add Spiral to Claude as an MCP server, and Claude can find and run MiniMax M3 (batch) like any built-in tool.

**How much does MiniMax M3 (batch) cost in Claude?**

Up to $0.004697 per run at the default inputs. $0.2912 per 1M input tokens, $1.144 per 1M output tokens. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Claude plan is separate.

**Do I need an OpenRouter account to use MiniMax M3 (batch) in Claude?**

No. Spiral runs MiniMax M3 (batch) on its own OpenRouter account. You sign in to Spiral once and never paste a key.

**How long does it take to set up MiniMax M3 (batch) in Claude?**

About two minutes: add Spiral to Claude, then keep the detected sign-in and click Add, then click Connect.

**Does Claude wait while MiniMax M3 (batch) runs?**

MiniMax M3 (batch) runs in the background. Claude starts the run, checks on it and gives you the text, json when it's ready.

**What is MiniMax M3 good at?**

MiniMax M3 is suited to bug fixing, frontend and backend development, performance optimization, and research that involves search. Give it a concrete deliverable, supporting evidence, and a way to check the result.

**How much context can MiniMax M3 handle?**

MiniMax documents a 1,000,000-token context window. That makes large document collections and repository analysis practical uses; ask for source references so you can check which material supports each conclusion.

**How long can MiniMax M3's output be?**

MiniMax's completion documentation specifies a maximum generation allowance of 524,288 tokens and recommends 131,072 tokens. Treat those as maker specifications; request a bounded deliverable such as a patch, a comparison table, or a report with a defined length.

**Which sampling settings does MiniMax recommend?**

The model card recommends temperature 1.0 and top_p 0.95. MiniMax's current documentation says its Anthropic-compatible interface fixes those values, so do not assume changing them will change the result.

**Where does MiniMax M3 still need careful checking?**

Difficult kernel optimization needs measured validation: MiniMax reports 28.8% on KernelBench Hard. For that work, require correctness checks and reproducible timing results before accepting a change.

