Muse Spark 1.3

4.55
If you only looked at the standard pricing for Muse Spark 1.3, you might think there wasn’t much to talk about. Input is still $1.25 per million tokens, output is still $4.25, and the headline specs haven’t changed much.
Advertisement 728 × 90
CompanyMeta
Context1M
Released2026-09
Updated2026-09-07

Muse Spark 1.3 Overview

Muse Spark 1.3 is Meta’s multimodal reasoning model released on September 2, 2026, less than a month after Muse Spark 1.2.

It accepts text, images, video, audio, and PDFs, with text-only output.

The model has a 1 million-token context window and supports up to 131,000 output tokens.

The big focus this time is pretty easy to spot: coding and Agent workflows.

Meta isn’t just pushing for better one-shot answers. A lot of the work in 1.3 seems to be about how the model behaves across a full task: how often it calls tools, how much context it uses, whether it knows when to ask for clarification, and how well it holds onto information deep into a long run.

Muse Spark 1.3 Pricing

PlanPriceDescription
Standard Input: $1.25 / 1M tokens; cached input: $0.15; output: $4.25 Data is not used for training. Rate limit: 3,000 requests/minute. Better fit for production workloads.

Both plans use the same Muse Spark 1.3 model, with the same 1 million-token context window and 131,000-token output limit.

There’s no free tier and no monthly subscription. Billing is usage-based.

The Contributor plan is dramatically cheaper, especially on output, but it comes with looser data-use terms and a much lower rate limit. For public code or personal experiments, that trade may be easy to accept. For proprietary code or customer data, it’s a different conversation.

Muse Spark 1.3 Key Features

1. Coding: Fewer Wasted Steps Matter More Than Longer Answers

The coding improvements in Muse Spark 1.3 are less about flashy demos and more about task efficiency.

Meta says that compared with 1.2, the model uses:

  • About 20% fewer tool calls
  • About 25% fewer tokens to complete the same tasks
  • Shorter responses with fewer unnecessary turns

That’s the part I’d pay attention to.

Real coding tasks rarely look like “write one function and stop.” A model may need to inspect files, find related code, call tools, make an edit, check the result, and then do another pass.

If two or three of those steps are pointless, the task still gets done, but it costs more and takes longer.

On DeepSWE v1.1, Muse Spark 1.3 scored 75.4, ahead of GPT 5.6 Sol at 73.0 and Opus 5 at 74.0.

On Terminal-Bench 2.1, it scored 88.8.

Those benchmark numbers put 1.3 in strong company, but I think the more useful part is the reduction in wasted work.

When an Agent is running a long coding task, one fewer tool call means one less wait. One fewer failed branch means one less chunk of context to process. That adds up quickly.

2. Agent Workflows: The Biggest Risk Is Being Wrong for Too Long

The Agent side of Muse Spark 1.3 is arguably more interesting than the raw coding scores.

Meta says the model is more likely to ask for clarification when a prompt is ambiguous instead of silently filling in the missing details.

It’s also supposed to stop more often when it hits a blocker and ask the user before taking important actions.

That sounds conservative, but that’s usually what you want from an Agent.

A chatbot misunderstanding one sentence is annoying. An Agent misunderstanding the task and then spending the next 15 steps editing files, calling tools, and building on that mistake is much worse.

Muse Spark 1.3 also shows a large jump in MRCR long-context testing:

  • 256K–512K: 98.5, versus 66.3 for 1.2
  • 512K–1M: 98.1, versus 55.5 for 1.2

That’s more useful than simply saying the model has a 1 million-token context window.

A huge context window only matters if the model can still find and use information that appeared hundreds of thousands of tokens earlier. Otherwise, it’s just a big number on the spec sheet.

3. Multimodal Input and a 1 Million-Token Context Window

Muse Spark 1.3 can take text, images, video, audio, and PDFs. Output is text only.

The context window is 1 million tokens, roughly around 750,000 English words.

For normal chat, that’s overkill.

For codebases, long documents, multi-file jobs, and extended Agent sessions, it’s much easier to see the point.

With a smaller context window, you often have to decide which files to include, trim code manually, and keep re-explaining the same project background.

A bigger window reduces some of that prep work.

I still wouldn’t read “1 million tokens” as “dump your entire repo in and forget about it.” Capacity and retrieval quality are two different things. The stronger MRCR scores are what make the context window more interesting here.

4. How Much Better Is It Than Muse Spark 1.2?

Muse Spark 1.2 launched on August 5. Muse Spark 1.3 arrived on September 2.

That’s a very short gap, so on paper it looks like a minor version bump.

The actual changes are more focused than that.

Terminal-Bench 2.1 moved from 82.9 to 88.8. Long-context performance improved sharply. The model also appears less likely to keep pushing through a bad assumption during long tasks.

I wouldn’t describe 1.3 as a model that suddenly got dramatically smarter.

A better way to put it is that it seems to waste less effort.

It calls fewer tools, uses fewer tokens, and is more willing to stop when the task is unclear. Those are small improvements in isolation, but they matter much more when the job runs for 20, 50, or 100 steps.

Most of the detailed performance numbers still come from Meta’s own testing. Artificial Analysis also points in the same general direction, but I’d still want to see more third-party results across different coding environments and Agent setups.

Summary

Muse Spark 1.3 doesn’t feel like a “bigger model, bigger numbers” release.

It looks more like Meta is trying to fix the annoying parts of long-running AI work.

Use fewer tools. Burn fewer tokens. Don’t keep charging ahead when the instructions are fuzzy. Hold onto information better when the context gets huge.

None of that sounds especially dramatic, but those are exactly the things that start to matter once an Agent is working for more than a few minutes.

The 1 million-token context window is useful for the same reason. It’s not especially exciting for basic chat, but it makes much more sense for codebases, large documents, and multi-step workflows.

There are still limits. Output tops out at 131,000 tokens, there’s no open-weight version, and some of the strongest performance claims still need broader independent testing.

As for Contributor, I’d treat it as an aggressive pricing option rather than the main reason to care about Muse Spark 1.3. It’s compelling for public code, personal projects, and low-sensitivity experiments. If you’re feeding it proprietary source code, customer information, or anything your company would hate to see reused, the cheap token price stops looking quite so cheap.

My read on Muse Spark 1.3 is pretty simple: it doesn’t seem much more eager to impress you. It just seems less likely to waste your time.

Comments (0)

Leave a comment

Advertisement 728 × 90

Meta Model Comparison

Model Context Pricing API Released Global Heat
Muse Spark 1.3
1M YES 2026-09
91/100
1M YES 2026-09
79/100
1M YES 2026-08
89/100
1M YES 2026-08
78/100
1M YES 2026-07
90/100

Similar Models

Llama 4 Scout
90
Meta open-source LLM allowing commercial use, local execution, and fine-tuning.
Meta
Muse Spark 1.1
90
Meta’s multimodal reasoning model optimized for AI Agents, coding, tool use, and complex task automation.
Meta
Muse Spark 1.2
89
Muse Spark 1.2 landed just a month after version 1.1. This release is more focused. Meta is putting more weight behind code generation, software engineering, and its new terminal-based agent, Muse Code.
Meta
Muse Spark 1.3 Contributor
79
The price is the story. Muse Spark 1.3 Contributor costs $0.20 per million output tokens, compared with $4.25 on the standard tier. The tradeoff is straightforward: prompts and outputs sent through Contributor may be used by Meta for model training
Meta
Muse Spark 1.2 Contributor
78
The first thing that caught my attention about Muse Spark 1.2 Contributor wasn’t the low price. It was the question behind it: why is Meta willing to make it this cheap? The answer is in the terms. If you let Meta use your interactions to improve its models, you get much cheaper API access.
Meta
Claude Opus 4.8
100
Claude Opus 4.8 is Anthropic's most powerful Opus-series model, featuring multimodal input, reasoning, and a 1M-token context window, excelling in complex reasoning and coding.
Anthropic
Claude Opus 5
100
Anthropic’s flagship Claude model built for complex reasoning, AI Agents, software development, and enterprise knowledge workflows.
Anthropic
GPT-5.5 Pro
100
I’ve been using GPT-5.5 Pro with a few colleagues for the past several weeks, mostly on the kinds of jobs where regular chatbots tend to fall apart: long documents, messy research questions, code debugging, and tasks that need more than one or two steps of reasoning.
OpenAI

Related News