[Integrations](https://spiral.ai/integrations) › [Devin](https://spiral.ai/integrations/devin) › [Llama 3.3 70B Instruct](https://spiral.ai/integrations/devin/llama-3-3-70b-instruct)

# How to Use Llama 3.3 70B Instruct in Devin

Use Llama 3.3 70B Instruct for multilingual conversations, document summaries, and coding assistance. It improves text performance over Llama 3.1 70B and supports a 128K context window.

|  |  |
| --- | --- |
| Model | Llama 3.3 70B Instruct |
| Available through | OpenRouter |
| Category | LLM |
| Price | Up to $0.003815 per run at the default inputs. $0.923 per 1M input tokens, $0.923 per 1M output tokens. |
| Returns | text, json |
| Runs | In the background |
| Setup | About 2 minutes, no API key |

## Prompts to try with Llama 3.3 70B Instruct in Devin

- Use Llama 3.3 70B Instruct to trace this project's authentication flow. Identify the entry points, session checks, and failure paths, citing file paths and function names. Keep the explanation under 600 words.
- Use Llama 3.3 70B Instruct to investigate the failing tests in this project. Explain the likely cause, propose the smallest relevant patch, and run the affected tests before reporting the result.
- Use Llama 3.3 70B Instruct to review the current diff for behavior changes. Report each concern with its file location, triggering input, and a test that would confirm it.
- Use Llama 3.3 70B Instruct to draft Spanish documentation for the project's public functions. Preserve code identifiers and explain parameters, return values, and errors using the implementation as evidence.
- Use Llama 3.3 70B Instruct to draft 12 fictional support conversations for this project's documented features. Include English and French examples, expected answers, and the documentation section supporting each answer.

## How to set up Llama 3.3 70B Instruct in Devin

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

### Use Spiral's MCP server

1. **Add a custom MCP**: Open Customize › MCPs. Click + and choose New MCP. Fill in:

2. **Sign in to Spiral and approve**: In Customize › MCPs, open Spiral and choose Connect. Spiral opens: sign in, pick the organization and click Authorize. No API key.

3. **Check it**: On Spiral's configuration page, click Test tools. Devin lists Spiral's tools.

### Use the Spiral skill

1. **Send Devin 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 Llama 3.3 70B Instruct**: 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": "meta-llama/llama-3.3-70b-instruct",
    "input": {
      "messages": [
        {
          "role": "system"
        }
      ]
    }
  }'
```

Then paste one of the prompts above: Devin finds Llama 3.3 70B Instruct on Spiral, shows the price and runs it.

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

## Llama 3.3 70B Instruct tools and prices

- [Meta: Llama 3.3 70B Instruct](https://spiral.ai/tools/openrouter/meta-llama/llama-3.3-70b-instruct) (OpenRouter): Up to $0.003815 per run at the default inputs. $0.923 per 1M input tokens, $0.923 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 Llama 3.3 70B Instruct in Devin?

Turn project code and documentation into explanations, proposed fixes, and test drafts with Llama 3.3. Give it a focused task and require evidence from the files it examines.

### Document summaries with evidence

Ask Llama 3.3 to condense a long report into decisions, risks, and open questions, with a supporting passage for each conclusion.

### Support replies in supported languages

Give Llama 3.3 the customer's message and relevant policy text. Request a reply in Spanish, French, Hindi, or another officially supported language, with unresolved questions flagged.

### Code explanations and proposed fixes

Use Llama 3.3 to explain a function, investigate a reported bug, and propose a patch with test cases. Run the tests before accepting the change.

### Meeting follow-up drafts

Give Llama 3.3 a meeting transcript and request decisions, action items, and unresolved questions. Ask it to leave owners and deadlines unspecified when the transcript omits them.

### Answers from supplied reference material

Ask Llama 3.3 to answer questions using a handbook or technical reference. Require a supporting section for each answer and an explicit acknowledgment when evidence is missing.

### Synthetic evaluation examples

Use Llama 3.3 to draft fictional support conversations or instruction-following examples for an evaluation set. Review labels and expected answers before using them.

### Tool-assisted research summaries

Have Llama 3.3 interpret results returned by search or lookup tools and draft an answer tied to those results. Keep missing evidence visible instead of filling gaps from memory.

## Overview

- [OpenRouter](https://spiral.ai/providers/openrouter)
- [All Devin guides](https://spiral.ai/integrations/devin)

## Llama 3.3 70B Instruct in other agents

- [Amp](https://spiral.ai/integrations/amp/llama-3-3-70b-instruct)
- [ChatGPT](https://spiral.ai/integrations/chatgpt/llama-3-3-70b-instruct)
- [Claude](https://spiral.ai/integrations/claude-desktop/llama-3-3-70b-instruct)
- [Claude Code](https://spiral.ai/integrations/claude-code/llama-3-3-70b-instruct)
- [Cline](https://spiral.ai/integrations/cline/llama-3-3-70b-instruct)
- [Codex](https://spiral.ai/integrations/codex/llama-3-3-70b-instruct)
- [Cursor](https://spiral.ai/integrations/cursor/llama-3-3-70b-instruct)
- [dots](https://spiral.ai/integrations/dots/llama-3-3-70b-instruct)
- [Factory Droid](https://spiral.ai/integrations/factory-droid/llama-3-3-70b-instruct)
- [Gemini](https://spiral.ai/integrations/gemini/llama-3-3-70b-instruct)
- [Gemini CLI](https://spiral.ai/integrations/gemini-cli/llama-3-3-70b-instruct)
- [Google Antigravity](https://spiral.ai/integrations/antigravity/llama-3-3-70b-instruct)
- [goose](https://spiral.ai/integrations/goose/llama-3-3-70b-instruct)
- [Grok](https://spiral.ai/integrations/grok/llama-3-3-70b-instruct)
- [Grok Build](https://spiral.ai/integrations/grok-build/llama-3-3-70b-instruct)
- [Hermes](https://spiral.ai/integrations/hermes/llama-3-3-70b-instruct)
- [Instinct](https://spiral.ai/integrations/instinct/llama-3-3-70b-instruct)
- [Manus](https://spiral.ai/integrations/manus/llama-3-3-70b-instruct)
- [Muse](https://spiral.ai/integrations/muse/llama-3-3-70b-instruct)
- [Muse Code](https://spiral.ai/integrations/muse-code/llama-3-3-70b-instruct)
- [n8n](https://spiral.ai/integrations/n8n/llama-3-3-70b-instruct)
- [Notion](https://spiral.ai/integrations/notion/llama-3-3-70b-instruct)
- [OpenClaw](https://spiral.ai/integrations/openclaw/llama-3-3-70b-instruct)
- [opencode](https://spiral.ai/integrations/opencode/llama-3-3-70b-instruct)
- [Perplexity](https://spiral.ai/integrations/perplexity/llama-3-3-70b-instruct)
- [Raycast](https://spiral.ai/integrations/raycast/llama-3-3-70b-instruct)
- [VS Code](https://spiral.ai/integrations/vscode/llama-3-3-70b-instruct)
- [Windsurf](https://spiral.ai/integrations/windsurf/llama-3-3-70b-instruct)
- [Zapier](https://spiral.ai/integrations/zapier/llama-3-3-70b-instruct)

## More Devin guides

- [Llama 3.1 Euryale 70B v2.2](https://spiral.ai/integrations/devin/llama-3-1-euryale-70b-v2-2)
- [Llama 4 Maverick](https://spiral.ai/integrations/devin/llama-4-maverick)
- [Llama 3.1 8B Instruct](https://spiral.ai/integrations/devin/llama-3-1-8b-instruct)
- [Llama 4 Scout](https://spiral.ai/integrations/devin/llama-4-scout)
- [Llama 3.1 70B Instruct](https://spiral.ai/integrations/devin/llama-3-1-70b-instruct)
- [Llama Guard 4 12B](https://spiral.ai/integrations/devin/llama-guard-4-12b)

## Questions about Llama 3.3 70B Instruct in Devin

**Can Devin use Llama 3.3 70B Instruct?**

Yes. Add Spiral to Devin as an MCP server, and Devin can find and run Llama 3.3 70B Instruct like any built-in tool.

**How much does Llama 3.3 70B Instruct cost in Devin?**

Up to $0.003815 per run at the default inputs. $0.923 per 1M input tokens, $0.923 per 1M output tokens. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Devin plan is separate.

**Do I need an OpenRouter account to use Llama 3.3 70B Instruct in Devin?**

No. Spiral runs Llama 3.3 70B Instruct on its own OpenRouter account. You sign in to Spiral once and never paste a key.

**How long does it take to set up Llama 3.3 70B Instruct in Devin?**

About two minutes: add a custom MCP, then sign in to Spiral and approve, then check it.

**Does Devin wait while Llama 3.3 70B Instruct runs?**

Llama 3.3 70B Instruct runs in the background. Devin starts the run, checks on it and gives you the text, json when it's ready.

**What is Llama 3.3 good at?**

Llama 3.3 suits multilingual conversations, document summaries, reasoning tasks, and coding assistance. Practical requests include summarizing a report, explaining a function, or drafting a support reply from supplied policies.

**Can Llama 3.3 read images or listen to audio?**

No. Llama 3.3 processes text, so provide a transcript or extracted text when your source is a recording or image.

**Which languages does Llama 3.3 support?**

Meta officially supports English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai. Specify the desired language and audience in your request.

**How much text can Llama 3.3 handle?**

Llama 3.3 has a 128K context window. For long documents, identify the sections to examine and request a focused answer with supporting passages.

**Does Llama 3.3 know current events?**

Its pretraining knowledge stops in December 2023. Supply current source material when asking about later events, and ask it to distinguish sourced statements from missing information.

