Meta: Llama 4 Maverick
meta-llama/llama-4-maverickGenerate and analyse text with Meta: Llama 4 Maverick from OpenRouter through one Spiral API key. It runs in the background and returns text and structured JSON. Pay as you go, with no subscription.
Pricing
Up to $0.005342 per run at the default inputs. $0.455 per 1M input tokens, $1.30 per 1M output tokens.
Limits
- Runs
- In the background
- Returns
- text, json
- Longest queue wait
- 1 hour
- Run deadline
- 30 minutes
- Longest run, end to end
- 225 minutes
- Largest inline input
- 1 MB
- Largest input file
- 2 GB
- Largest output
- 5 GB
Use Meta: Llama 4 Maverick in your AI agent
Connect Spiral to your agent once, then ask for Meta: Llama 4 Maverick by name. Each guide has the setup and prompts.
Input
| Field | Kind | Default | Description |
|---|---|---|---|
messages required | JSON | — | The conversation so far, oldest first. |
seed | number (whole) | — | |
stop | JSON | — | |
tools | JSON | — | |
top_k | number (whole) 0 to … | — | |
top_p | number 0 to 1 | — | |
logprobs | true or false | — | |
max_tokens | number (whole) 1 to 32768 | — | The most output tokens to generate. |
temperature | number 0 to 2 | — | |
tool_choice | JSON | — | |
top_logprobs | number (whole) 0 to 20 | — | |
response_format | JSON | — | |
presence_penalty | number -2 to 2 | — | |
frequency_penalty | number -2 to 2 | — | |
repetition_penalty | number 0 to 2 | — |
How to use Meta: Llama 4 Maverick
From your AI agent, the command line or your own code. One Spiral account runs it every way.
Use Spiral's MCP server
- Add Spiral's MCP server
Any MCP client takes https://api.spiral.ai/mcp with your API key as a Bearer token. In Claude Code:
claude mcp add spiral --transport http https://api.spiral.ai/mcp \ --header "Authorization: Bearer $SPIRAL_API_KEY"
- Ask for Meta: Llama 4 Maverick
Your agent finds it on Spiral, shows the price and runs it.
Use Meta: Llama 4 Maverick from OpenRouter on Spiral.
Use the Spiral skill
- Send your agent this message
Your agent reads Spiral's skill and sets itself up.
set up https://spiral.ai/SKILL.md
- Sign in or paste a key
Approve Spiral in the browser, or give the agent a key from the console.
- Ask for Meta: Llama 4 Maverick
Your agent finds it on Spiral, shows the price and runs it.
Use Meta: Llama 4 Maverick from OpenRouter on Spiral.
Use the Spiral CLI
- Install and sign in
Installs the Spiral CLI, signs you in and adds its skill to your coding agents.
npx -y spiral-cli@latest init
- Run Meta: Llama 4 Maverick
Waits for the result and saves the run and its files under .spiral/runs.
spiral runs create \ --provider openrouter \ --model meta-llama/llama-4-maverick \ --input '{ "messages": [ { "role": "system" } ] }' \ --wait --save
Call the Spiral API
- Create an API key
In the console, under API keys.
- Run Meta: Llama 4 Maverick
One request; the result comes back by polling or a webhook.
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-4-maverick", "input": { "messages": [ { "role": "system" } ] } }'
More on Meta: Llama 4 Maverick
Provider
Frequently asked questions
How much does Meta: Llama 4 Maverick cost?
Up to $0.005342 per run at the default inputs. $0.455 per 1M input tokens, $1.30 per 1M output tokens. It is paid from your prepaid Spiral balance as you use it, with no subscription.
How long does Meta: Llama 4 Maverick take to run?
It runs in the background: the request returns a run id at once, and you poll `GET /v1/runs/{id}` or name a webhook to hear when it finishes. A run can take at most 225 minutes, end to end.
What input does Meta: Llama 4 Maverick take?
It needs `messages`. It also takes 14 optional fields: `seed`, `stop`, `tools` and 11 more. Each field's type, allowed values and default are listed on this page.
What does Meta: Llama 4 Maverick return?
A finished run returns text and json output. Each output up to 5 GB is stored as a copy; a larger output comes back as the provider's own link where the provider gives one, and otherwise that output is not delivered and you are not charged for it; a run with nothing delivered fails as too large, without a charge.
What are the limits of Meta: Llama 4 Maverick?
Longest queue wait: 1 hour; largest inline input: 1 MB; largest input file: 2 GB.
How do I use Meta: Llama 4 Maverick through an API?
Send `POST /v1/runs` with `"provider": "openrouter"`, `"model": "meta-llama/llama-4-maverick"` and its input, using one Spiral API key.
How do I see what a run of Meta: Llama 4 Maverick will cost before I start it?
Send the same request to `POST /v1/estimate`: it prices that exact input without running it.
What happens if a run of Meta: Llama 4 Maverick fails?
Only for work the provider billed. A run that fails before the provider does billable work is not charged, and its hold goes back to your balance; a failure the provider bills for is charged for the work the provider billed, at the tool's price. Each failed run's error says why it failed and what it was charged.
Do I need an account with OpenRouter to use Meta: Llama 4 Maverick?
No. Spiral runs it on its own OpenRouter account and charges your Spiral balance.
Can my AI agent use Meta: Llama 4 Maverick?
Yes. Any agent connected to Spiral's MCP server, CLI or agent skill runs it with the same API key, and there are guides for using it from Amp, ChatGPT, Claude, Claude Code, Cline, Codex, Cursor, Devin, dots, Factory Droid, Gemini, Gemini CLI, Google Antigravity, goose, Grok, Grok Build, Hermes, Instinct, Manus, Muse, Muse Code, n8n, Notion, OpenClaw, opencode, Perplexity, Raycast, VS Code, Windsurf and Zapier.
Can I cap what an agent spends on Meta: Llama 4 Maverick?
Yes. Give each agent its own API key with a spend cap: a run whose estimate would pass the cap is refused before it starts.

dots
