What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Meta’s Contributor tier for Muse Spark 1.3 is cheaper than Standard because you give something in return. Meta’s Model API pricing documentation describes the tier as “Heavily discounted token pricing in exchange for permission to use your prompts and completions to train future Meta models.” For coding-agent and API work, the prompts and completions are the code, instructions, and generated output your tools send. Before you choose the cheaper tier, you need to decide whether that material is suitable for training use.
The rate card, side by side
Meta’s pricing page lists two tiers for Muse Spark 1.3. The table below uses the list rates shown on that page, retrieved October 7, 2026. The ratio column is calculated from those rates and is not a Meta figure.
As an Amazon Associate I earn from qualifying purchases.
| Line item (per million tokens) | Standard | Contributor | Standard ÷ Contributor |
|---|---|---|---|
| Input, uncached | $1.25 | $0.10 | 12.5× |
| Cached input | $0.15 | $0.002 | 75× |
| Output | $4.25 | $0.20 | 21.25× |
| Published training terms | Prompts and completions are not used to train Meta models | Permission to use prompts and completions to train future Meta models | Not applicable |
The gap is widest on cached input and narrowest on uncached input. If your coding agent repeatedly resends a large, stable context such as a repository summary or system instructions, the cached line is the one that will drive your bill.
What the discount is paid for
Meta frames the two tiers as a direct trade. Standard is the tier where Meta says prompts and completions are not used to train Meta models. Contributor is the tier where you grant that permission in exchange for lower rates. That distinction is what the pricing page establishes, and it is the reason the discount exists.
#1 Best Overall
The model listing page uses shorter labels: “Used to improve our products” for Contributor and “Not used to improve our products” for Standard. The pricing page wording is more precise, because it names training of future models specifically. Neither source, as reviewed, sets out retention periods, data-processing terms, or other handling rules. Those are separate questions from whether material may be used for training, and you should not assume the training permission answers them.
What a coding agent actually sends
Meta announced Muse Code in beta on August 5, 2026. It is a terminal coding agent that was powered at launch by Muse Spark 1.2, and Meta says it handles software-engineering tasks across repositories, including planning, writing code, and validating results. An agent that works across a repository does not send only the question you type. The model request can include whatever the agent reads in order to do the job.
Cobalt’s third-party setup guide gives one concrete example. In its Muse Code task, the traffic can include repository source, diffs, test output, and the files the agent reads. That describes how one harness behaved in that guide. It is not a universal description of Muse Code, and other tools, accounts, or configurations may send less or send it differently.
The sources reviewed do not establish which tier a Muse Code session bills against by default. Check your account and harness settings to confirm the model ID and tier actually in use before assuming the Contributor rates apply.
What the numbers look like for a workload
The rates above are token prices, not an estimate of any real coding task. To show how the gap plays out, the table uses a hypothetical month of 10 million input tokens and 1 million output tokens. The second scenario assumes 8 million of those input tokens are served from cache. These volumes are illustrative assumptions, not a typical-workload benchmark, and the totals cover only the token rates listed.
| Scenario (hypothetical monthly volume) | Standard | Contributor |
|---|---|---|
| A: 10M uncached input + 1M output | $16.75 (10 × $1.25 + 1 × $4.25) | $1.20 (10 × $0.10 + 1 × $0.20) |
| B: 2M uncached input + 8M cached input + 1M output | $7.95 (2 × $1.25 + 8 × $0.15 + 1 × $4.25) | $0.42 (2 × $0.10 + 8 × $0.002 + 1 × $0.20, rounded from $0.416) |
Scenario B shows why caching matters. Once most input is cache-served, the Standard-to-Contributor difference is driven by output and the small uncached remainder, and the absolute dollar gap shrinks. Substitute your own token counts from your provider usage reporting before drawing budget conclusions.
Rank #4
Where the “20x cheaper” figure comes from
An independent pricing guide poses the question “Why is the Contributor tier 20x cheaper?” That phrasing is a rounded shorthand, and the guide’s content is based on Muse Spark 1.2 rates. Against the current Muse Spark 1.3 list rates, the ratio is 12.5× for uncached input, 75× for cached input, and 21.25× for output. Only the output line comes close to 20×. If you repeat the comparison, name the line item, and use the current rates.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →A decision framework for coding workloads
- Classify the material. Public open-source code is a different case from proprietary, client-owned, or regulated code. Anything containing credentials, personal data, or customer records should not go to Contributor without written approval from your security or legal owner.
- Map what the harness sends. List the repository files, diffs, test output, logs, and environment output your agent can read. Run it against a non-sensitive repository and check what actually appears in the requests before using it on real work.
- Match the material to the tier. If the material is acceptable under the published Contributor terms, the lower rates are a real option. If it is not, use Standard for that work instead.
- Narrow the context. Exclude secret files, environment files, and customer data from what the agent can read. A tighter context also lowers token volume on the Standard tier.
- Get the integration assessed. The published tier wording describes Meta’s terms, not your contract or the specific tool you use. Ask the security or legal owner to review the actual integration and provider terms, not just the price page.
These steps are editorial recommendations based on the documented tier distinction and the kinds of input a coding agent can send. They are not additional Meta policy statements.
Best Value
Keeping these figures current
Rates and model versions change. Muse Spark 1.2 and 1.3 have separate model IDs and tier rates, so a rate you read for one may not apply to the other. Meta’s official model page listed Muse Spark 1.3 with a 1-million-token context window when checked on October 7, 2026. Before you set a budget or approve a workflow, check the current Meta Model API pricing page and the model ID your tool is configured to call.
- Confirm the exact model ID in your harness or API configuration.
- Confirm the current input, cached-input, and output rates for that tier.
- Re-read the training-permission wording for the tier you use.
- Repeat these checks whenever the tool, the model, or the contract changes.
Within those limits, the decision is simple: the Contributor rates are only worth taking for work you are willing to let Meta use for training.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →
