Muse Spark 1.3

Language Meta Partly verified Released Sep 2, 2026

In Input modalities: TextImageVideoPDFAudio Out Output modalities: Text
Reasoning· minimal, low, medium, high, xhigh, max
Tool use
Structured output

Prices

1 offering · prices as listed

Meta AI API Global · USD muse-spark-1.3

Conditions/ 1M tokens	Input	Cache read	Output
Standard	$1.25	$0.15	$4.25

Prices from Meta Model API pricing and rate limits (opens in a new tab) (checked Nov 8, 2026)

Standard: the price a request with no conditions pays (the headline price elsewhere on the site).

Details

Open weights: —
Family: muse-spark
Model ID

Not stated on Meta’s pages: max input, reasoning can be off, max thinking budget, built-in tools, knowledge cutoff, aliases.

Cited scores

From third-party leaderboards, not measured by us

Epoch Capabilities Index 158

#23 of 118 listed models · CI 155.9–160.2
LMArena Text 1,496

#10 of 163 listed models · max effort · CI 1,491–1,501 · 25,563 votes
Epoch AI · Chess Puzzles 37%

#40 of 85 listed models · max effort · ±4.85

Also listed: xhigh effort 34%
Epoch AI · FrontierMath Tier 4 v2 46.3%

#23 of 65 listed models · max effort · ±7.88

Also listed: xhigh effort 41.5%

Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Oct 28 – Nov 5, 2026
LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Nov 8, 2026
EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · License · as of Oct 28, 2026
LiveBench (White et al.), release 2026-06-25, livebench.ai, licensed CC BY-SA 4.0. This board, adapted from it, is shared under CC BY-SA 4.0. · License · as of Jun 25, 2026
Terminal-Bench 4.0 leaderboard (Marten, Shaw et al., Terminal-Bench; harbor-framework/terminal-bench on GitHub), licensed Apache-2.0. · License · as of Sep 29, 2026
changes we made

Provider-reported scores

Published by Meta, under the conditions it states

Benchmark	Setting	Score
Terminal-Bench 2.1	max reasoning effort, native coding harness, 89 tasks, pass@1	88.8%
	xhigh reasoning effort, native coding harness, 89 tasks, pass@1	89.2%
DeepSWE 1.1	max reasoning effort, mini-swe-agent, task pass rate	75.4%
MRCR v2	8-needle, 256K–512K tokens, max reasoning effort	98.5%
	8-needle, 256K–512K tokens, xhigh reasoning effort	97.6%
	8-needle, 512K–1M tokens, max reasoning effort	98.1%
	8-needle, 512K–1M tokens, xhigh reasoning effort	93.1%
DeepSearchQA	F1, browser tools, max reasoning effort	90.3%
	F1, browser tools, xhigh reasoning effort	89.4%

News

All news →

Sep 2, 2026Release Meta releases Muse Spark 1.3 On September 2, 2026, Meta released Muse Spark 1.3, including a max-reasoning option, on Muse Code and Meta Model API. It targets stronger agentic and coding work and more reliable long instructions.

Changes

All changes →

Sep 26, 2026 New language model

Sources

Muse Spark 1.3 Evaluation Methodology (PDF, Meta) (opens in a new tab) research.meta.ai · checked Sep 26, 2026
Meta Model API models (opens in a new tab) dev.meta.ai · checked Nov 8, 2026
Meta Model API: reasoning (opens in a new tab) dev.meta.ai · checked Nov 8, 2026
Meta Model API pricing and rate limits (opens in a new tab) dev.meta.ai · checked Nov 8, 2026
Research | Meta AI Research (opens in a new tab) research.meta.ai · checked Nov 3, 2026
models.dev: muse-spark-1.3 (community data, MIT) (opens in a new tab) github.com · checked Sep 30, 2026

Meta’s official policies: data use, privacy, terms →

Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data

Meta · United States · ai.meta.com

Snapshots, the same family and other models from this provider