TokenSave
Muse Spark 1.3 API Pricing: Meta's Model vs Claude, GPT-6 and Gemini
Jonhisking · October 9, 2026
Meta's Muse Spark 1.3 is available on the Meta Model API at $1.25 per million input tokens and $4.25 per million output tokens, with cached input at $0.15. There is no long-context surcharge on its 1M-token context window. Meta also offers a much cheaper "contributor" version that lets Meta train on your data. Here is what Muse Spark 1.3 costs next to Claude Sonnet 5.5, GPT-6 Sol, Gemini 3.1 Pro, DeepSeek and Grok.
Muse Spark 1.3 API pricing
Per million tokens, from Meta's official pricing page:
| Model name | Input | Cached input | Output |
|---|---|---|---|
muse-spark-1.3 (standard) |
$1.25 | $0.15 | $4.25 |
muse-spark-1.3-contributor |
$0.10 | $0.002 | $0.20 |
- Same price for 1.1, 1.2 and 1.3. Meta says the standard versions share one price, so upgrading costs nothing extra.
- No long-context premium. You pay the same rate whether the 1M-token window is nearly empty or almost full.
- Contributor tier: Meta uses your prompts and completions to train future models. It is cheap, but do not send private or customer data to it.
- Rate limits (standard): 3,000 requests and 4,000,000 tokens per minute per team.
Muse Spark 1.3 vs other models
List prices per million tokens:
| Model | Input | Output |
|---|---|---|
| Muse Spark 1.3 | $1.25 | $4.25 |
| DeepSeek V4 Pro | $1.32 | $3.96 |
| Grok 4.7 | $2.00 | $6.00 |
| GPT-6 Sol | $2.00 | $10.00 |
| Claude Sonnet 5.5 | $2.00 | $10.00 |
| Gemini 3.1 Pro | $2.00 | $12.00 |
What it costs on real work
A typical request here is 2,000 input tokens and 500 output tokens. Prices include each model's tokenizer difference on the same English text (Claude uses about 30% more tokens; Meta's tokenizer is not public, so we treat it like GPT's).
| Task | Muse Spark 1.3 | Claude Sonnet 5.5 | GPT-6 Sol | Gemini 3.1 Pro | DeepSeek V4 Pro | Grok 4.7 |
|---|---|---|---|---|---|---|
| One typical request | $0.0046 | $0.0117 | $0.0090 | $0.0095 | $0.0046 | $0.0070 |
| 10,000 requests a month | $46 | $117 | $90 | $95 | $46 | $70 |
- On this request, Claude Sonnet 5.5 costs 153% more than Muse Spark 1.3, GPT-6 Sol 95% more, and Gemini 3.1 Pro 105% more.
- DeepSeek V4 Pro costs about the same. Grok 4.7 costs 51% more.
What Meta says it is good at
According to Meta, Muse Spark 1.3 is trained for long, multi-step agent work and tuned for coding, with "fewer unnecessary turns and cleaner output." Meta also says it reads images, video and documents natively. These are Meta's claims; test it on your own tasks before you switch.
Strengths and weaknesses
Strengths
- Price: $1.25 / $4.25 per million tokens, below Claude Sonnet 5.5 and GPT-6 Sol.
- Long context: 1M tokens with no surcharge. On Meta's long-context recall test (MRCR, 512K–1M tokens) it scores 98.1, against 73.8 for GPT-5.6 Sol.
- Coding: Meta reports DeepSWE v1.1 at 75.4 (up from 55.0 on 1.2) and Terminal-Bench 2.1 at 88.8, the same as GPT-5.6 Sol. Meta's engineers saw about 20% fewer tool calls and 25% fewer tokens than 1.2 on the same tasks.
- Independent ranking: at its max setting it scores 62 on the Artificial Analysis Intelligence Index, 6th of 636 models.
Weaknesses
- Verbose: Artificial Analysis measured 120 million output tokens to finish its index, against a median of 72 million. That is about 67% more output, so your real bill can be higher than the per-token price suggests. Measure it on your own prompts.
- Not first on agent tasks: in Meta's own table, Claude Opus 5 is slightly ahead on agent benchmarks such as OSWorld 2.0 (68.3 vs 66.9), and GPT-5.6 Sol leads on search and instruction following.
- Max mode came later: the max reasoning mode behind the top score was held back for safety testing at launch; the generally available xhigh setting scores 61.
- No visible reasoning: it does not show its thinking, which some users find harder to debug.
- Closed model, one provider: no open weights, and Meta is the only host.
Should you switch?
- On Claude Sonnet 5.5, GPT-6 Sol or Gemini 3.1 Pro for cost: Muse Spark 1.3 is a lot cheaper per request. Run a small test on your real prompts and compare quality.
- On DeepSeek V4 Pro: the price is about the same, so choose on quality, speed and where your data is processed.
- For non-sensitive bulk jobs: the contributor tier ($0.10 / $0.20) is one of the cheapest options, as long as you are fine with Meta training on that data.
- Using it in Cursor: Cursor lists Muse Spark 1.3 at the same $1.25 / $4.25 in its "other models" pool.
Check your own prompt
Paste your prompt into the token counter to see its cost on Muse Spark 1.3 and 30+ other models, or compare Muse Spark 1.3 vs Claude Sonnet 5.5 side by side.
Prices checked October 9, 2026. Check Meta's pricing page before you commit.
Sources
- Meta Model API: Pricing and rate limits
- Meta Model API: Muse Spark 1.3
- Cursor: Models & Pricing
- Artificial Analysis / eesel: Muse Spark 1.3 review
- DataCamp: Muse Spark 1.3
More articles
- ChatGPT Pro 500 vs 200 vs 100: Price, Limits and Which to Pick (2026)
- ChatGPT Free vs Paid (2026): What You Actually Get for $0, $8, $20 and $100
- I Ran My Released Game Through My AI Game Cost Calculator: Prediction vs Reality
- What's Trending on Roblox Right Now (October 2026): 8 Game Types
- How to Make a Roblox Game with AI (2026): Steps, Tools, Cost
Paste a prompt to see its tokens and cost on Muse Spark 1.3, GPT-6, Claude, Gemini and 30+ models.
Open the token counter →