When was Muse Spark Open released?

Meta released Muse Spark Open on 15 Oct 2026. It is available through the Opper gateway.

Does Muse Spark Open support zero data retention?

Yes. Muse Spark Open is available with zero data retention on Fireworks through the Opper gateway.

Where is Muse Spark Open hosted?

See the per-route table for Muse Spark Open hosting locations and residency details.

Is Muse Spark Open trained on prompts and outputs?

No. Every hosting route for Muse Spark Open is configured so customer prompts and outputs are not used to train models by default.

Is a GDPR DPA available for Muse Spark Open?

Yes. A GDPR Data Processing Agreement is available on at least one route serving Muse Spark Open. Contact hello@opper.ai to receive it.

Muse Spark Open

by Meta

Released 43 days after Muse Spark 1.3 Contributor · Announcement · All Meta releases

Muse Spark Open is the open-weight release of Muse Spark, Meta's agentic and coding model, published on October 15, 2026. It is a sparse mixture-of-experts transformer with 816B total parameters and about 42B active per token, so per-token compute tracks the active parameters while serving still loads the full weights. Of its 62 decoder layers, 46 use sliding-window attention over the most recent 2,048 tokens and 16 use full attention, which keeps the key-value cache small across a 1,048,576-token context. The model is natively multimodal: built-in perception encoders take images, audio and video alongside text, and it answers in text. Meta built it for long-horizon tasks, complex software engineering and multi-step reasoning, with tool calling for agent loops. It always reasons before it answers, and a reasoning effort from low to xhigh (the default) sets how much, trading answer quality against latency and tokens. Meta publishes the weights in BF16 and in NVFP4, which keeps most weights in 4-bit, under the Muse Model License with limited commercial terms. The model was trained on data from more than 100 languages, and Meta's model card reports its evaluations at xhigh effort.

Call Muse Spark Open on Opper with the OpenAI SDK. Sign up without a credit card. Get started

Get your API key Try in chat Compare

Key info

Input

Output

Features

Context window

Max output

—

Input price

$1.25 /1M$0.15 cache read

Output price

$4.25 /1M

Released

15 Oct 2026

Web search

Via Opper

Zero data retention on pay-as-you-go
No training by default
GDPR DPA available

Available routes

Muse Spark Open runs on 2 different routes through the Opper gateway. Compare residency, zero data retention and training posture at a glance, with full data-handling detail per route below.

Provider	Region	Zero data retention	Training	Input	Output	Cache read
Fireworks	Multi		No	$1.25	$4.25	$0.15
Nebius	Multi		No	$1.25	$4.25	Input price

Live routes and prices as JSON: api.opper.ai/v3/models?q=muse-spark-open

Uptime and availability

Muse Spark Open runs on 2 independently monitored routes through the Opper gateway. A direct integration leaves you on one of them, the gateway serves whichever is healthy.

100%effective uptime, last 30 days

Averaged across these routes on their own, a single provider reached 93.96% over the same window.

Daily status through the gateway, which serves each request from whichever route is healthy.

2 monitored providers serve Muse Spark Open.

Measured over the last 30 days from each provider's official status feed via StatusGator. Refreshed hourly. See uptime for every provider Opper monitors.

Data handling per route

Each route hosting Muse Spark Open has its own privacy posture, residency, and GDPR terms. Postures are maintained by Opper with a last-verification timestamp.

Fireworks — Multi-region

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. GLOBAL; SCCs; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: On by default, keeps nothing
Subprocessor access: Subprocessors may read content
GDPR DPA: DPA available
Transfer mechanism: SCCs

Nebius — Multi-region

Content is retained: provider logging. No training on customer data. GLOBAL; SCCs; DPA available.

Zero data retention: Retained: provider logging.
Training: No training on customer data.
Logging: Other
Abuse monitoring: No classifier
GDPR DPA: DPA available
Transfer mechanism: SCCs

Best for

Where Muse Spark Open ranks among the models on Opper with a score for each use case.

Legal#15 of 4139.6

Legal tasks and reasoningSee the Legal ranking →Healthcare#16 of 3839.1

Healthcare and medicalSee the Healthcare ranking →Finance#18 of 4142.0

Finance and accountingSee the Finance ranking →Coding7.8

FrontierCode 2.0 and Terminal-Bench 4.0See the Coding ranking →Agents#41 of 8628.9

Terminal-Bench 4.0 and AutomationBenchSee the Agents ranking →

Benchmarks

Independent benchmark scores: the Intelligence Index and the individual evaluations behind it, grouped by skill.

Global rank#31 of 506 LLMs

TierEfficient

Output speed177 tok/s

Intelligence Index24.9

Coding

FrontierCode 2.0

Terminal-Bench 4.0

SciCode

58%

Agentic & knowledge work

AutomationBench

50%

GDPval

45%

GDP.pdf

19%

Briefcase (Elo)1278

Science & reasoning

Terminal-Bench Science

CritPt (hard)

30%

HLE Diamond

26%

Humanity's Last Exam

44%

GPQA Diamond

89%

Long-context reasoning

78%

Knowledge

Omniscience25.8

MMMU-Pro

81%

By domain

Finance & accounting

42.0

Legal

39.6

Healthcare & medical

39.1

Engineering

35.1

Economics

40.6

Strategy & operations

43.4

See full leaderboard →Benchmarks via Artificial Analysis · View on AA

Get started

Call Muse Spark Open through the Opper gateway with one API key. Opper is drop-in compatible with the OpenAI, Anthropic and Google AI SDKs.

Muse Spark Open is a premium model on Opper. Sign up needs no credit card: you get an API key straight away and the free models work in the playground and the API. Add a card to use premium models, pay-as-you-go with no minimum.

Set it up with your agent

Copy this and paste it into a coding agent like Claude Code, Cursor or Codex and it'll wire up Opper for you.

Or call it directly

1import OpenAI from "openai";
2
3const client = new OpenAI({
4  apiKey: process.env.OPPER_API_KEY,
5  baseURL: "https://api.opper.ai/v3/compat",
6});
7
8const completion = await client.chat.completions.create({
9  model: "muse-spark-open",
10  messages: [{ role: "user", content: "Hello" }],
11});
12console.log(completion.choices[0].message.content);