Muse Code Free Tier: No Free Plan, and the Contributor Tier Is Not the Free One

Muse Code has no free tier — no free plan, no trial, no free credit allocation and no rate-limited gratis access on Meta's own product page, subscriptions page or changelog, all three of which we read on 14 October 2026. That one sentence is the whole answer for anyone who arrived from muse code free tier, is muse code free, muse code free, does muse code have a free tier or muse code trial. The product is Muse Code, Meta's coding agent for the terminal and CI; the models it drives are Meta's Muse Spark 1.3, Muse Spark 1.2 and Muse Spark 1.1; and neither the harness nor any of those three models is free to run. If you typed muse spark 1.3 free for the model rather than the harness, that is a different question with the same kind of answer, and the boundary between the two is set out below.

Read this as a standing page, not as news. Muse Code went into beta in August 2026 and left beta on 31 August 2026, and those dates are here to place the product in time rather than to announce anything. What the page exists for is the question readers are typing now. Our own Search Console measurement for the 28 days to 12 October 2026 recorded 52 impressions and no clicks on the three free-shaped query rows — muse code free tier at 32, is muse code free at 12 and muse code free at 8 — plus three further rows on the page-level view: does muse code have a free tier at 5, muse spark 1.3 free at 5 and muse code trial at 1. All four of those spellings are the same question, and only the first of the four questions they stand for is a flat yes-or-no.

What Muse Code costs today, read on 14 October 2026

Meta's Muse Code product page lists four subscription plans today, not three: Everyday Usage at $5.00 per month, High Usage at $15.00 per month, Power Usage at $50.00 per month, and Maximum Usage at $200.00 per month. The fourth plan is new since our own Muse Code reference page was written on 28 September 2026, which carried three plans at $5.00, $15.00 and $50.00. Meta's page wins, so the figure has moved and you are reading the moved one.

Here is the current list as Meta writes it, read today:

• Everyday Usage — $5.00/month; access to Muse Spark, Muse Wonder and Muse Image; 10–50 prompts every 5 hours, including image and video uploads; voice mode; web search

• High Usage — $15.00/month; everything in the Everyday Usage plan; 5× more usage than Everyday; more multimodal inputs

• Power Usage — $50.00/month; everything in the High Usage plan; access to the most powerful Muse models, including Muse Wonder Full; 20× more usage than Everyday; early access to new features; higher file uploads

• Maximum Usage — $200.00/month; everything in the Power Usage plan; 80× more usage than Everyday

Two things about that list bear on the free-tier question. First, every plan Meta describes is described by what it includes — prompts per window, multimodal inputs, a usage multiple — and none of the four is framed as a trial, an introductory offer or a free starting point. Maximum Usage is written as 80× the Everyday plan, a multiple of a paid plan, not a step down from something free. Second, the price does not move when you change the Contributor setting. Meta's own help-centre wording is that turning Contributor mode off "changes how much usage you get from your plan. It doesn't change which plan you're on." There is no combination of settings that makes the monthly charge zero.

What the Contributor tier actually is — in the two places it exists

This is the misreading the page exists to fix. "Contributor" in Meta's Muse documentation names two different mechanisms, and neither one is a free tier. If you came here thinking Contributor is the free one, the mistake is understandable and the correction is short.

In Muse Code, Contributor is a mode, not a plan. Meta's subscriptions page states it directly: "Muse Code subscriptions default to Contributor mode. In Contributor mode, you get more usage from your plan in exchange for letting Meta use your content — the prompts, code, files, and context you send, plus the model responses you get back — to train future Meta models." The current changelog line, release 1.4.6, adds the availability rule: Contributor mode "is subscribers-only and routes each model family to its served contributor model; without an active subscription or a served contributor model the toggle stays hidden and requests go out unchanged." A toggle that is subscribers-only, hidden without a served contributor model, and switched in /settings is not a free tier and cannot reasonably be mistaken for one. Note also that Contributor mode is not a discount you get automatically: 1.4.6 made it on by default for subscribers, and Meta's help page says you can turn it off at any time, after which your future content is not used for training.

On Meta Model API, Contributor is a model variant with its own rate card. Meta's pricing page names the two Contributor variants of Muse Spark — muse-spark-1.3-contributor and muse-spark-1.2-contributor — at $0.10 per million input tokens, $0.20 per million output and $0.002 per million cached input, against the Standard rows for muse-spark-1.3, muse-spark-1.2 and muse-spark-1.1 at $1.25 input, $4.25 output and $0.15 cached. Meta's product page labels the Contributor rows "Used to improve our products." and the Standard rows "Not used to improve our products." That is roughly twelve times cheaper on input and a little over twenty-one times cheaper on output — a genuine discount, and still not a free tier, because $0.10 is not $0.00.

What the reader gives up for that price is a data clause, not a feature. The prompts, code, files and context you send, plus the responses that come back, may be used to train future Meta models. Meta's Contributor-tier help page is explicit about the responsibilities that follow: "You must not submit sensitive, confidential, or personal information to the Discounted Services", and if your use case involves personal information, Meta's instruction is to use the Standard services instead. Contributor mode is not available to accounts with Zero Data Retention enabled, so the cheapest rate and the strongest retention posture are mutually exclusive by design. Subscribers must be 18 or the age of majority in their country, and anyone shipping an integrated product on the Contributor tier has to maintain a privacy policy telling their own users what is collected and whether it feeds advertising.

Two further details matter and both are easy to miss. First, on Meta Model API the variant is not automatic — you get the discounted rate by calling the -contributor model id explicitly, so a request sent to muse-spark-1.3 is billed at Standard rates whether or not you intend it. Second, the discount is bought partly with throughput: Meta's rate card today gives the Contributor variants 150 requests per minute per team against 3,000 for the Standard rows. Contributor is a data-governance decision before it is a budget decision, and it is never a free tier.

Where Muse Code's billing stops and the model's rate card starts

muse spark 1.3 free is not the same question as muse code free, and readers merge them because Meta's own marketing does, by describing the harness and the model as one thing. They are two products with two bills.

Muse Code is the harness. It is a native muse binary you run inside a project directory; it plans, edits files and runs commands, with approvals and an OS sandbox active from the first run. It is not a model, and Meta does not sell it by the token. Muse Spark 1.3, Muse Spark 1.2 and Muse Spark 1.1 are the models it drives, and those Meta sells per token on Meta Model API. When you buy a Muse Code subscription you are buying usage of the harness: Meta's help page says the subscription "works only through the Muse Code CLI while signed in with your Meta Model API account", and that it applies to "the Muse Code API key that is automatically connected in the Muse Code CLI onboarding process. This credential is for use with Muse Code only. Any additional API keys you create under your Meta Model API account will be billed through pay-as-you-go." When you buy Muse Spark tokens you are buying inference on a key of your own.

So the answer splits cleanly along that seam. There is no free tier for the harness, and there is no free rate for the models. muse-spark-1.3 is $1.25 per million input tokens and $4.25 per million output on the Standard tier today, and muse-spark-1.3-contributor is $0.10 and $0.20. Zero is not on either card. A reader who searched for a free Muse Spark 1.3 has been handed the same flat no as a reader who searched for a free Muse Code, for a different reason: one is a subscription that does not exist, the other is a rate that does not exist.

The "platform free-tier credits" line, handled honestly

The phrase matters because it has been quoted as evidence of a Muse Code free tier, and it is not one. Our own Muse Code reference page recorded it in a Muse Voice Transcribe note on Meta's pricing and rate-limits page, where it described the pricing of zero-data-retention transcription rather than anything a Muse Code user can spend. Muse Voice Transcribe is a separate product on Meta Model API, billed by the hour of audio processed rather than by token, and the note sits under its own heading, not under Muse Code's. If you find that line quoted as evidence of a Muse Code free tier, it is being read out of context.

Say the next part plainly too. On our reading today the sentence is not in Meta's copy at all. We read the pricing and rate-limits page, the Muse Voice Transcribe cookbook page, the speech-to-text guide and the voice API reference on 14 October 2026, and swept every URL in Meta's own sitemap for the word "free" and for the phrases "free tier", "free trial", "free credit" and "free plan" — 404 pages in total, of which 240 are documentation, help-centre, resource or product pages. Every occurrence of "free" on those pages is ordinary English: free-form context, free-text answers, dependency-free code, CC0 game assets. The live Muse Voice Transcribe note now reads "$0.18 per hour", "Streaming and non-streaming transcription are priced the same", "Zero data retention (ZDR) is priced at parity with Standard mode" and "Contributor mode isn't available for Muse Voice Transcribe" — and carries no free-credit sentence. So the honest position is two-part: the line, where it existed, was never about Muse Code, and it is not on the page today either. We are not going to print it as though it were a Muse Code fact, and we are not going to print it as though it were a trick. Where Meta is silent we say Meta is silent, and on free credits for a Muse Code user Meta is silent today.

What OrcaRouter serves from this family

OrcaRouter is a single endpoint covering 200+ models at each provider's list rate with 0% markup, so the availability question here is checkable rather than a matter of policy. Checked against our live public model list today: the harness itself is not something we route — Muse Code is a coding agent rather than a model, so meta/muse-code is not in our catalogue and will not appear there. From Meta's Muse line, exactly two entries are served: meta/muse-spark-1.2 and meta/muse-spark-1.1, both at the same rates Meta publishes for them. meta/muse-spark-1.3 is not served, and neither Contributor variant is served. Read that as a description of today's catalogue and nothing more.

The practical reading is narrow and worth stating precisely, because it is one of the four questions readers arrive with. Two of the Muse models are reachable through us on one key, at Meta's own list price passed through — and neither is free, here or anywhere else. The harness's own numbers — what it costs per month, which CLI release is current, what Meta claims for it — belong to our Muse Code reference page, and the hands-on walkthrough of the terminal agent belongs to our Muse Code terminal coding agent page. Both are live, and this page stops where its own question stops.

What you can run today without paying Meta anything

Not nothing, and not a Muse Code tier: Muse Glimmer, Meta's open-weight multimodal model — the "30B" in its name rounds a card figure of 29.6 billion parameters including the vision encoder — distilled from Muse Spark and released under the Apache License 2.0. Meta publishes the weights on Hugging Face across four repositories — the bf16 safetensors checkpoint, quantised GGUF builds, a DFlash draft checkpoint and prebuilt ExecuTorch exports — and Meta's own get-the-model page says so outright: everything runs offline once the weights land, with no API key and no gated access. Meta's GGUF card lists its Q4_K_M build at 16.8 GB and describes that size as fitting 24 GB of VRAM, which means a machine that will never buy a Muse Code subscription can still run a Meta model on its own hardware, with no Meta account and no monthly charge.

Be exact about the limits of that answer, because it is not the same product. Muse Glimmer is a model, not a coding agent. Meta's own model card gives it 29.6 billion parameters including the vision encoder, and a sequence length of 131,072 tokens against the 1,048,576-token envelope Meta's hosted Muse Spark models carry. It also does not take audio, which the card states outright. And it is a different deployment posture: your weights, your runtime, your data. It does not replace Muse Code and it does not make Muse Code free. What it does mean is that the answer to "is there anything from Meta I can run without paying Meta" is yes, and the thing has a name.

The other half of the honest answer is what "free" means for Muse Code itself. Installing it costs nothing — Meta's one-line installer is curl -fsSL https://dev.meta.ai/install.sh | sh on macOS and Linux, and it puts a native muse binary on your path. Running it is where the bill starts, because the binary needs a credential before it can reach a model: either a subscription bought through Accounts Center, or an API key billed pay-as-you-go. There is no third path in which the agent runs against Meta's hosted models at Meta's expense. One adjacent thing not to count either: Meta's Playground, the web UI for trying models without writing code, is not a free sandbox — Meta's own help page states that requests made in the Playground count toward your team's usage and rate limits the same as API calls do.

So: no free plan, no trial, no credits; a Contributor mode that trades training rights for usage and a Contributor rate card that starts at $0.10 rather than $0.00; four subscription plans from $5.00 to $200.00 a month; and one genuine route to running a Meta model at no cost, which is Muse Glimmer's Apache-2.0 weights on your own machine. Everything above was read from Meta's own pages on 14 October 2026, and where Meta says nothing we have said so instead of filling the gap.

The harness's own numbers — what it costs per month, which CLI release is current, what Meta claims for it — belong to our Muse Code reference page and the hands-on walkthrough of the terminal agent belongs to our Muse Code terminal coding agent page. Both are live, and this page stops where its own question stops.

Answers like this one are also where the failure modes show up: the documented AI incidents and agent failures around models in this class are collected in the OrcaRouter AI Incident Archive , where each entry carries its own sources, severity, confidence grade and disputed flag.