[Integrations](https://spiral.ai/integrations) › [Claude Code](https://spiral.ai/integrations/claude-code) › [LFM2.5-24B-A2B](https://spiral.ai/integrations/claude-code/lfm2-5-24b-a2b)

# How to Use LFM2.5-24B-A2B in Claude Code

Use LFM2.5-24B-A2B for text-based agents that follow detailed instructions and call tools across multiple steps. Its 131,072-token context supports long conversations and documents, while built-in reasoning stays on for every answer.

|  |  |
| --- | --- |
| Model | LFM2.5-24B-A2B |
| Available through | OpenRouter |
| Category | LLM |
| Price | Up to $0.00373 per run at the default inputs. $0.065 per 1M input tokens, $0.91 per 1M output tokens. |
| Returns | text, json |
| Runs | In the background |
| Setup | About 2 minutes, no API key |

## Prompts to try with LFM2.5-24B-A2B in Claude Code

- Use LFM2.5-24B-A2B to investigate the first failing test in this project. Trace the relevant code, make the smallest justified fix, and rerun the affected tests; report the commands and results.
- Use LFM2.5-24B-A2B to compare the README's configuration instructions with the implementation. List mismatches with file and line references, then propose corrections grounded in the code.
- Use LFM2.5-24B-A2B to review the current diff for missing error handling. Follow each changed call to its implementation and report only findings supported by a concrete failure path.
- Use LFM2.5-24B-A2B to turn the requirements in this project's issue notes into an implementation plan. Identify dependencies, affected files, and a verification step for each requirement.
- Use LFM2.5-24B-A2B with temperature 0.1 and top-k 50 to check this project's Spanish and French interface text against the English strings. Preserve interpolation tokens and report meaning changes with file references.

## How to set up LFM2.5-24B-A2B in Claude Code

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

### Use Spiral's MCP server

1. **Add the hosted server**: This adds Spiral to the current project. Add `--scope user` to use it in every project.

```bash
claude mcp add --transport http spiral https://api.spiral.ai/mcp
```

2. **Sign in**: Start `claude`, type `/mcp`, choose spiral and Authenticate. Sign in to Spiral in the browser and click Authorize. No API key.

### Use the Spiral skill

1. **Send Claude Code 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 LFM2.5-24B-A2B**: 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": "liquid/lfm-2.5-24b-a2b",
    "input": {
      "messages": [
        {
          "role": "system"
        }
      ]
    }
  }'
```

Then paste one of the prompts above: Claude Code finds LFM2.5-24B-A2B on Spiral, shows the price and runs it.

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

## LFM2.5-24B-A2B tools and prices

- [LiquidAI: LFM2.5-24B-A2B](https://spiral.ai/tools/openrouter/liquid/lfm-2.5-24b-a2b) (OpenRouter): Up to $0.00373 per run at the default inputs. $0.065 per 1M input tokens, $0.91 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 LFM2.5-24B-A2B in Claude Code?

Use LFM2.5-24B-A2B to trace project code, work through a focused fix, and check the result with tools. Give it a bounded task and explicit acceptance criteria, then require test evidence.

### Research briefs with sources

Use LFM2.5-24B-A2B to search for evidence, compare findings, and write a brief with links attached to factual claims. Ask it to flag unanswered questions.

### Support answers grounded in policy

Have LFM2.5-24B-A2B retrieve the relevant policy passages, reconcile them with the customer's request, and draft a response that identifies missing information.

### Long document comparisons

Give LFM2.5-24B-A2B the text of two specifications or proposals. Request a comparison of requirements, contradictions, and open decisions with passage references.

### Small code fixes with verification

Ask LFM2.5-24B-A2B to inspect a failing test, trace the relevant code, and propose a focused patch. Require test results before accepting the fix.

### Multilingual request triage

Use LFM2.5-24B-A2B to summarize incoming text in supported languages and assign categories from a written rubric. Review ambiguous classifications.

### Daily operational digests

Have LFM2.5-24B-A2B compare current tool results with the previous report, identify changes, and produce a short digest with evidence for each exception.

### Plans with dependencies

Give LFM2.5-24B-A2B a goal, constraints, and available resources. Request an ordered plan that makes dependencies and unresolved assumptions explicit.

### Multistep calculation checks

Ask LFM2.5-24B-A2B to compare quantities, schedules, or capacity assumptions from supplied text. Require stated units and an independent calculation check.

## Overview

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

## LFM2.5-24B-A2B in other agents

- [Amp](https://spiral.ai/integrations/amp/lfm2-5-24b-a2b)
- [ChatGPT](https://spiral.ai/integrations/chatgpt/lfm2-5-24b-a2b)
- [Claude](https://spiral.ai/integrations/claude-desktop/lfm2-5-24b-a2b)
- [Cline](https://spiral.ai/integrations/cline/lfm2-5-24b-a2b)
- [Codex](https://spiral.ai/integrations/codex/lfm2-5-24b-a2b)
- [Cursor](https://spiral.ai/integrations/cursor/lfm2-5-24b-a2b)
- [Devin](https://spiral.ai/integrations/devin/lfm2-5-24b-a2b)
- [dots](https://spiral.ai/integrations/dots/lfm2-5-24b-a2b)
- [Factory Droid](https://spiral.ai/integrations/factory-droid/lfm2-5-24b-a2b)
- [Gemini](https://spiral.ai/integrations/gemini/lfm2-5-24b-a2b)
- [Gemini CLI](https://spiral.ai/integrations/gemini-cli/lfm2-5-24b-a2b)
- [Google Antigravity](https://spiral.ai/integrations/antigravity/lfm2-5-24b-a2b)
- [goose](https://spiral.ai/integrations/goose/lfm2-5-24b-a2b)
- [Grok](https://spiral.ai/integrations/grok/lfm2-5-24b-a2b)
- [Grok Build](https://spiral.ai/integrations/grok-build/lfm2-5-24b-a2b)
- [Hermes](https://spiral.ai/integrations/hermes/lfm2-5-24b-a2b)
- [Instinct](https://spiral.ai/integrations/instinct/lfm2-5-24b-a2b)
- [Manus](https://spiral.ai/integrations/manus/lfm2-5-24b-a2b)
- [Muse](https://spiral.ai/integrations/muse/lfm2-5-24b-a2b)
- [Muse Code](https://spiral.ai/integrations/muse-code/lfm2-5-24b-a2b)
- [n8n](https://spiral.ai/integrations/n8n/lfm2-5-24b-a2b)
- [Notion](https://spiral.ai/integrations/notion/lfm2-5-24b-a2b)
- [OpenClaw](https://spiral.ai/integrations/openclaw/lfm2-5-24b-a2b)
- [opencode](https://spiral.ai/integrations/opencode/lfm2-5-24b-a2b)
- [Perplexity](https://spiral.ai/integrations/perplexity/lfm2-5-24b-a2b)
- [Raycast](https://spiral.ai/integrations/raycast/lfm2-5-24b-a2b)
- [VS Code](https://spiral.ai/integrations/vscode/lfm2-5-24b-a2b)
- [Windsurf](https://spiral.ai/integrations/windsurf/lfm2-5-24b-a2b)
- [Zapier](https://spiral.ai/integrations/zapier/lfm2-5-24b-a2b)

## More Claude Code guides

- [Kimi Latest](https://spiral.ai/integrations/claude-code/kimi-latest)
- [LFM3-1.7B](https://spiral.ai/integrations/claude-code/lfm3-1-7b)
- [Kimi K3.1 Pro (batch)](https://spiral.ai/integrations/claude-code/kimi-k3-1-pro-batch)
- [Ling 3.0 Flash](https://spiral.ai/integrations/claude-code/ling-3-0-flash)
- [Kimi K3.1 Pro](https://spiral.ai/integrations/claude-code/kimi-k3-1-pro)
- [Ling 3.0 Flash VL](https://spiral.ai/integrations/claude-code/ling-3-0-flash-vl)

## Questions about LFM2.5-24B-A2B in Claude Code

**Can Claude Code use LFM2.5-24B-A2B?**

Yes. Add Spiral to Claude Code as an MCP server, and Claude Code can find and run LFM2.5-24B-A2B like any built-in tool.

**How much does LFM2.5-24B-A2B cost in Claude Code?**

Up to $0.00373 per run at the default inputs. $0.065 per 1M input tokens, $0.91 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 Code plan is separate.

**Do I need an OpenRouter account to use LFM2.5-24B-A2B in Claude Code?**

No. Spiral runs LFM2.5-24B-A2B on its own OpenRouter account. You sign in to Spiral once and never paste a key.

**How long does it take to set up LFM2.5-24B-A2B in Claude Code?**

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

**Does Claude Code wait while LFM2.5-24B-A2B runs?**

LFM2.5-24B-A2B runs in the background. Claude Code starts the run, checks on it and gives you the text, json when it's ready.

**What is LFM2.5-24B-A2B good at?**

Tool use and following detailed instructions are its clearest strengths. Consider it for tasks that combine several lookups, decisions, or checks before producing an answer.

**Where does LFM2.5-24B-A2B fall short?**

Liquid AI reports that larger comparison models retain an edge on knowledge-heavy and coding benchmarks. For factual research, pair it with retrieved sources or web search; for code changes, check the result with tests.

**How much context can LFM2.5-24B-A2B handle?**

The documented context length is 131,072 tokens. It is a text-only model, so use document text or transcripts for document analysis.

**Which languages does LFM2.5-24B-A2B support?**

Liquid AI lists 16 languages: English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish, Vietnamese, Thai, Indonesian, Hindi, Russian, and Polish. The list does not establish equal accuracy across languages, so check examples from your actual task.

**Can I turn off reasoning in LFM2.5-24B-A2B?**

No supported thinking-off switch is documented. Reasoning and the final answer share the generation token limit, so leave room for both.

