Model

Muse Spark 1.1 (xhigh)

CLOSED

Meta · Muse Spark family · released Jul 9, 2026

First flagship from Meta Superintelligence Labs (led by Alexandr Wang), replacing the open Llama line with a proprietary model; base Muse Spark launched 2026-04-08, this is the 1.1 update from 2026-07-09, evaluated here at the 'xhigh' reasoning-effort tier. Meta has said it hopes to open-source future versions but this release is closed/API-only.

ReasoningCodingVisionFunction callingTool useAgentic

2605.1

Elo · rank #10

Parameters

Undisclosed

Active params

—

Context

1M tokens

Architecture

Proprietary multimodal reasoning model (undisclosed architecture), Meta Superintelligence Labs' first model family, update to April 2026 Muse Spark

License

Proprietary

Languages

—

API price (in/out)

$1.25 / $4.25

Modalities

text · vision

Benchmark results

Bar shows position within the tracked field; marker = field best

CharXivVision88.4%#1

best: this model · 88.4%

CyberGymCoding59.0%#5

best: GPT-5.6 · 84.5%

GDPval-AAAgents1376#21

best: Claude Fable 5 · 1932

GPQA DiamondReasoning90.0%#28

best: GPT-5.6 · 94.6%

Humanity's Last ExamReasoning45.0%#11

best: Claude Sonnet 5 · 57.4%

LiveCodeBenchCoding85.9%#20

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MMLU-ProKnowledge88.7%#5

best: Claude Fable 5 · 91.5%

MMMUVision86.6%#4

best: Claude Fable 5 · 89.3%

OSWorld-VerifiedAgents80.8%#4

best: Claude Fable 5 · 85.0%

SWE-bench ProCoding61.5%#8

best: Claude Fable 5 · 80.0%

SWE-bench VerifiedCoding82.0%#8

best: Claude Fable 5 · 95.0%

Terminal-Bench 2.0Coding78.0%#10

best: GPT-5.6 · 88.8%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

$1.25

Output / M tok

$4.25

Muse Spark family

Elo progression across releases

Muse Spark 1.1 (xhigh)Jul 20262605.1

Resources

DOCShttps://ai.meta.com/APIAvailable via Meta API

API price $1.25/$4.25 · each benchmark row carries its own source badge (see methodology)