Everything about Muse Spark 1.3

Overview

Meta released Muse Spark 1.3 on September 2, 2026. It has a reasoning mode: it can "think" before answering, which helps with multi-step problems but uses more tokens, and therefore more money. Its weights are not published, so it only runs remotely, through the Meta API or partner hosts (19 listed on models.dev). It is a very recent release.

We file it under budget models (24 models in this category on the site).

How much does Muse Spark 1.3 cost?

Muse Spark 1.3 costs $0.10 per million input tokens and $0.20 per million output tokens (developer's price, checked on October 22, 2026).

In practice, an agent task that reads 20,000 tokens and writes 2,000 costs about $0.002, or $2.4 for 1,000 such tasks.

That is 5.6 times cheaper than the median output price of budget models ($1.13).

Tokens read back from cache cost $0.002 per million, which helps an agent that keeps resending the same context.

To estimate your own budget, use the agent cost calculator.

What do independent benchmarks say?

According to results published by Epoch AI (checked on October 22, 2026), Muse Spark 1.3 has a capability index of 82.7 out of 100, computed on 3 benchmarks from our basket.

With only 3 benchmarks reported, it does not enter the public ranking, which requires at least 4 results, so its index should be read as partial.

• OTIS Mock AIME (competition mathematics problems): 99.2% — the best tracked model, GPT-5.5 Pro, reaches 100%
• FrontierMath (very hard research-level mathematics): 74.4% — the best tracked model, GPT-6 Astra, reaches 93.7%
• APEX-Agents (professional tasks carried out by an agent): 58.6% — the best tracked model, Gemini 4 Argon, reaches 82.4%

Context and capabilities

Its context window reaches 1,048,576 tokens, roughly 1,573 pages of text at 500 words per page: enough to load a whole book or a mid-sized codebase. A single answer can run to 131,072 tokens.

It accepts as input: text, image, video, PDF and audio.

It can call tools (functions, APIs) and return structured JSON output, which makes it a possible foundation for an AI agent.

What is it good for?

Given its category, Muse Spark 1.3 is mainly aimed at high volumes where every cent counts: sorting, information extraction, summaries, simple assistants.

Its very large context window makes it worth considering for analysing large sets of documents in a single request.

Its very low cost (under half a cent for our sample task) means it can be called thousands of times a day.

Before adopting it, test it on your own examples: benchmarks measure standard tasks, not your job.

What could replace it?

In the same category, the models closest to Muse Spark 1.3 are:

• GPT-6 Luna (OpenAI, index 80.3) — see the comparison
• GPT-5.6 Luna (OpenAI, index 79.6) — see the comparison
• Qwen3.6 Flash (Alibaba Cloud (Qwen), index 64.5) — see the comparison
• Gemini 3.1 Flash Lite (Google DeepMind, index 64.3) — see the comparison