Meta's Model API sells Muse Spark 1.3 on pay-as-you-go per-token billing, with versions 1.2 and 1.1 still callable at the same Standard rate. Below are the rates, what typical workloads cost, and how cached input and the discounted Contributor mode change the bill.
Muse Spark 1.3 costs $1.25 per 1M input tokens and $4.25 per 1M output tokens, with cached input at $0.15 and a 1M context window. Meta charges no long-context premium, so a near-full prompt is billed at the same per-token rate as a short one. Web search grounding is billed separately, per query, on top of tokens.
| Model | Tier | Input / 1M | Output / 1M | Cache hit / 1M | Context | Notes |
|---|---|---|---|---|---|---|
| Muse Spark 1.3 | Mid-tier | $1.25 | $4.25 | $0.15 | 1M | Meta's first paid model, agentic; 1.3 succeeds 1.2 at same price (Sep 2, 2026 efficiency update) |
| Model | Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens |
|---|---|---|---|
| Muse Spark 1.3 | $42.00 | $40.90 | $1.68 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
| Model | Provider | Tier | Input / 1M | Output / 1M | Context |
|---|---|---|---|---|---|
| Muse Spark 1.3 | Meta | Mid-tier | $1.25 | $4.25 | 1M |
| Mistral Large 3 | Mistral | Mid-tier | $0.50 | $1.50 | 256K |
| DeepSeek V4-Pro | DeepSeek | Mid-tier | $0.66 | $1.98 | 1M |
| Mistral Large 4 | Mistral | Mid-tier | $0.70 | $2.10 | 1M |
| Grok 4.3 | SpaceXAI | Mid-tier | $1.25 | $2.50 | 1M |
Meta's cheapest model per tier (bold) against the four cheapest same-tier models from other providers, by output price.
Meta runs a second price list, Contributor mode (model IDs ending in -contributor, offered for Muse Spark 1.3 and 1.2), at a steep discount in exchange for permission to use your prompts, files and responses to train future Meta models. That trade rules it out for customer data, regulated content or anything under NDA, but it can make sense for synthetic-data generation, public-web summarisation or internal experiments. Contributor models also get a much lower request ceiling — 150 requests per minute versus 3,000 on Standard — so they suit back-office batch jobs rather than user-facing traffic. All tables on this page use Standard rates.
Cached input is billed at $0.15 instead of $1.25 per 1M tokens when the start of your prompt matches a cached prefix. Keep system prompts, tool definitions and long reference documents at the front of each request and byte-identical between calls, put the user-specific part last, and check the cached-token count in each response's usage data to confirm hits. Meta's rate sheet lists only a cached-input price, with no separate cache-write line.
Limits apply per team, not per API key: 3,000 requests and 4 million tokens per minute for each Standard Muse Spark model, counted separately per model. The API accepts OpenAI Responses, OpenAI-compatible Chat Completions and Anthropic-compatible Messages request formats, so existing OpenAI or Anthropic SDK code usually needs only a new base URL, key and model ID. Parallel tool calls, streamed tool-call arguments and JSON-schema structured outputs are supported.
At $4.25 per 1M output tokens, Muse Spark 1.3 is priced between low-cost options such as DeepSeek V4-Pro ($1.98) and mid-tier models like Claude Sonnet 5.5 ($10.00) or GPT-6.1 Sol ($10.00), and close to Z.ai's GLM-5.3 ($4.40). Meta also sells Muse Wonder (muse-wonder-1.0), a larger model aimed at complex software engineering, at a higher rate that this site doesn't track yet.
Standard Muse Spark 1.3 is $1.25 per 1M input tokens, $0.15 per 1M cached input tokens and $4.25 per 1M output tokens, billed pay-as-you-go with no minimum or upfront commitment.
No. Meta lists muse-spark-1.3, 1.2 and 1.1 at the same Standard price, so there is no cost reason to stay on an older version. Muse Spark 1.1 is Standard-only, with no Contributor variant.
A discounted price tier in which Meta may use your prompts, files and responses to train future models. It has lower rate limits (150 requests per minute) and isn't offered for Muse Wonder.
No. Meta states there is no long-context premium: the per-token rate is the same however much of the 1M window you fill.
Yes. Meta describes the Model API as drop-in compatible with the OpenAI SDK, the Anthropic SDK and OpenAI-compatible agent CLIs.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.