Skip to content
llmwaves

Muse Spark 1.2

Meta · released Aug 5, 2026

ProprietaryReasoningTool useStructured outputVisionAudio inVideo in

Muse Spark 1.2 is a reasoning model from Meta, designed for complex agentic tasks. It accepts text, images, video, audio, and PDF documents, returns text, and offers a 1M-token context...

Specification

Context window
1.05M
Max output
Knowledge cutoff
Not stated
Parameters
Undisclosed
Licence
Proprietary
Serving providers
1
Moderated
Yes
Uptime
100.0%

Intelligence

56.8

98th percentile

Coding

72.2

Coding Index

Agentic

Agentic Index

Output speed

125 t/s

Median across providers

Latency

4.47s

Time to first token

Cost per task

$0.40

Estimated

Benchmarks

Where the score comes from

The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.

Evaluation scores

Percentage correct · higher is better

  • GPQA Diamond
    90.4%
  • AA-LCR (long context)
    83.3%
  • SciCode
    56.4%
  • Humanity's Last Exam
    45.5%
  • τ²-bench (Telecom)
    34.8%

An evaluation missing from this list was not run for this model — it is not a zero.

View as table
EvaluationScore
GPQA Diamond90.4%
AA-LCR (long context)83.3%
SciCode56.4%
Humanity's Last Exam45.5%
τ²-bench (Telecom)34.8%

Against its peers

Intelligence Index · this model highlighted, nearest peers in grey

Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.

View as table
ModelIntelligence
Qwen3.8 Max58.1
Claude Opus 4.857.3
Muse Spark 1.256.8
GPT-5.6 Terra56.6
GPT-5.556.3
Grok 4.555.8
Claude Sonnet 555.3
Claude Opus 4.755.0

Percentile among all indexed models

Intelligence98th
Coding93th
Arena Elo94th

Pricing

What it costs to run

List prices per million tokens, plus what one representative task works out to.

List price

Input / 1M tokens
$1.25
Output / 1M tokens
$4.25
Cached input / 1M
$0.15
Blended 3:1
$2

One task, estimated

$0.40

Input tokens
50,000
Output tokens
80,000
Profile
Reasoning

Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).

Serving providers

Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.

Arena

Head-to-head generation quality

Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.

Overall Elo
1,322
Win rate
55.9%
Strongest at
Game development
Tournaments
102

Elo by category

Dot position on a 1,300–1,325 scale · Elo has no meaningful zero

1,3001,325

Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.

View as table
CategoryEloWin rate
Game development132255.9%
Websites131855.7%