Muse Spark Contributor API: $0.10/$0.20 and the Training Licence

Muse Spark 1.3 Contributor and Muse Spark 1.2 Contributor are Meta routes, live as muse-spark-1.3-contributor and muse-spark-1.2-contributor. The list price is $0.10 per 1M input tokens and $0.20 per 1M output tokens.

That is not a sale. Meta's standard Muse Spark rate is $1.25 / $4.25. The gap is what the training licence is worth. Prompts and completions you send to these IDs may be used to train future Meta models. RouterPlex does not train on your data; Meta may, on this route only.

Muse Spark Contributor API pricing #

RouterPlex does not sell the standard editions, and it does not reprice Contributor up to $1.25 / $4.25. The cheap rate belongs to the user who granted the licence.

A $5 prepaid balance buys roughly 50 million input tokens, or 25 million output tokens, if a workload used only one category. That math is irrelevant if the prompts are confidential.

The training licence is the product #

Three rules, copied from the catalog note and the privacy policy:

1. Do not send confidential, personal, or client-owned material. Once the prompt leaves the gateway, RouterPlex cannot retrieve it from Meta.
2. Never a default, never a fallback. A request reaches Contributor only when the model field names muse-spark-1.2-contributor or muse-spark-1.3-contributor.
3. Meta may limit availability to certain regions.

Every other model on this catalog stays on the no-training rule. See Privacy.

Tool calling is not supported #

Both Contributor IDs are tool_call: false on RouterPlex. The upstream serves function calls in a shape standard OpenAI clients do not read (a second choice, not choices[0]). Advertising tools would be a lie, so the catalog does not.

If you need tools, pick another model on the same key — GLM-5.3, Qwen3.8 Flash, Grok 4.6.

Cost per agent turn #

One 40,000-token prompt returning 2,000 tokens, no cache:

The arithmetic: 40,000 × $0.10 / 1M + 2,000 × $0.20 / 1M = $0.004 + $0.0004 = $0.0044. The dollars are not the expensive part.

Muse Spark Contributor specifications #

Call Muse Spark Contributor with RouterPlex #

Only use this if you have read the training licence and still want the ID.

curl https://api.routerplex.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "muse-spark-1.3-contributor",
    "messages": [
      {"role": "user", "content": "Summarise this public README in five bullets."}
    ]
  }'

import os
from openai import OpenAI
 
client = OpenAI(
    api_key=os.environ["ROUTERPLEX_API_KEY"],
    base_url="https://api.routerplex.com/v1",
)
 
response = client.chat.completions.create(
    model="muse-spark-1.3-contributor",
    messages=[{"role": "user", "content": "Rewrite this public changelog for humans."}],
)
 
print(response.choices[0].message.content)

Put Contributor on its own key with a hard budget, so a default model string elsewhere in the app cannot wander onto it. The live Muse Spark 1.3 Contributor page is the number this route will actually bill.