ModelPriceWatch.com
Last scan 2026-09-20 Models tracked 263 Providers 34 Cheapest paid Granite 4.0 H Micro $0.017/Mtok in Every price links to its source

Muse Spark 1.2

by Meta

Current flagship mid tier

Today's price · per 1M tokens

Input

$1.25

per 1M tokens

Output

$4.25

per 1M tokens

Blended

$2.00

blended $/1M — 3:1 weighted input:output

Cached input

$0.150

12% of input — prompt caching

How this price is scoped: Meta Model API standard-tier list price, per 1M tokens, with no long-context premium — the tier on which Meta states prompts and completions are not used for training. A cheaper contributor tier ($0.10 input / $0.20 output per 1M) trades training rights and a far lower rate limit (100 vs 3,000 requests/min) for the discount, so it is not the tracked list price. Search grounding is billed separately per 1,000 queries and is not included in the token rate.

Source: official Meta pricing · read MODELPRICEWATCH.COM · 2026-09-20

Price receipt

We read Meta's own pricing page on and found muse-spark-1.2 listed with a price on the same tier — that read is where the number above comes from. We keep the page text we read; its content hash is 849e200423.

Overview

The previous Muse Spark version on the Meta Model API, superseded by Muse Spark 1.3 on 2026-09-02 but still sold at the same rate — a multimodal agentic model with a 1M-token context that takes text, image, video, audio and PDF input, retrained for coding, whole-repository work and computer use. $1.25/$4.25 per 1M tokens, with cached input at $0.15 — the same list price as Muse Spark 1.1 and 1.3. Like 1.1 it is API-only: there are no open weights. Meta still recommends 1.2 over 1.3 for audio input, where 1.3's support is not yet complete.

Capabilities

struck through = not supported
Input 4/4
Text Image Audio Video
Output
Text
Features 5/9
Prompt caching Reasoning Coding Fast inference Long context Open weights Multimodal Web search Realtime

Benchmark performance

accuracy % · higher is better
Avg benchmark score62%
Perf per $/Mtok31
Terminal-Bench 2.1
80%
Humanity's Last Exam
44%
Independent composite scores
  • AA Intelligence Index: 39.8 (Artificial Analysis composite across reasoning, knowledge and coding evals)
  • AA-Omniscience Index: 27.2 (−100–100)
  • GDPval-AA v2: 1615 ELO (Real-world work tasks, human baseline = 1000)
  • LMSYS Chatbot Arena: 1499.6 ELO (Human preference)
How it stacks up
  • Ranks #55 of 84 benchmarked models by average score
  • Ranks #22 of 181 comparably-measured models by percentile score, across 6 independent measurements
  • Ranks #70 of 181 comparably-tested models by normalized performance per dollar
  • Strongest at Terminal-Bench 2.1 — 80%, #16 of 41

Source: Artificial Analysis, LMSYS Chatbot Arena (UC Berkeley) · updated · See full rankings →

Specifications

Provider
Meta
Context window
1M tokens
Modality
text, image, video, audio
Parameters
Proprietary
Open source
No — proprietary
Released
Status
Current
Last updated
Tags
flagshipmultimodalagenticreasoningcodingcaching

Availability verified: listed on Meta's own page