Patchwire

Meta · Released Aug 5, 2026

Muse Spark 1.2

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

ReasoningVision
  • Top 17%SimpleBench
  • Top 19%SciCode
  • Top 25%SimpleQA Verified
  • Top 32%LMCA
Compare in showdown → OpenRouter ↗ All Meta models
Released
Aug 5 2026
Context window
1M 944K out
Input / 1M tokens
$1.25
Output / 1M tokens
$4.25

Where to run it

Providers

Only Meta serves Muse Spark 1.2 on OpenRouter right now.

ProviderInputOutputCache readContextMax outputUptimevs. best
Meta $1.25 $4.25 $0.151M944K100%24h Cheapest

Value

Price vs. performance

7 models score higher for the same price or lesson FrontierSWE, e.g. Claude Haiku 5.5.
12.0%29.8%47.7%65.5% $0.01$0.10$1.00$10.00 Gemini 4 Argon (Google) — 55.0% · $2.00/1M in Claude Haiku 5.5 (Anthropic) — 43.8% · $0.10/1M in Claude Sonnet 5.5 (Anthropic) — 61.9% · $2.00/1M in Claude Opus 5.5 (Anthropic) — 62.3% · $4.00/1M in Grok 4.7 (xAI) — 29.5% · $2.00/1M in GPT-6 Astra (OpenAI) — 65.5% · $10.00/1M in Qwen3.8 Max (0902) (Qwen) — 17.8% · $2.00/1M in Gemini 3.8 Flash (Google) — 19.6% · $0.82/1M in Claude Fable 5.1 (Anthropic) — 56.3% · $10.00/1M in GLM 5.3 Flash (Z.ai) — 18.1% · $0.047/1M in GLM 5.3 (Z.ai) — 30.2% · $0.079/1M in Gemini 3.7 Flash (Google) — 20.3% · $0.75/1M in Grok 4.6 (xAI) — 25.3% · $2.00/1M in Claude Opus 5 (Anthropic) — 52.0% · $5.00/1M in Kimi K3 (Moonshot AI) — 25.9% · $0.24/1M in GPT-5.6 Sol (OpenAI) — 32.2% · $4.00/1M in Claude Fable 5 (Anthropic) — 47.0% · $10.00/1M in Muse Spark 1.3 (Meta) — 25.8% · $1.25/1M in Muse Spark 1.2 (Meta) — 12.0% · $1.25/1M in Muse Spark 1.2

Muse Spark 1.2 Other Meta Other labs Value frontier · FrontierSWE vs. input price per 1M tokens (log scale) · Drag to pan, pinch or Ctrl + scroll to zoom, double-click to zoom in.

Benchmarks

12 benchmark scores

BenchmarkScoreRankLeader
FrontierSWESoftware engineering · Ultra-long-horizon engineering tasks with a 20-hour budget each 12.0% #19of 19 GPT-6 Astra65.5%
DeepSWESoftware engineering · Original long-horizon tasks across active open-source repos 54.9% #16of 26 GPT-6 Astra74.1%
SciCodeSoftware engineering · Research-coding problems from the natural sciences 57.4% #17of 93 Claude Opus 5.566.9%
APEX-AgentsAgents & computer use · Long professional tasks in banking, consulting and law 45.4% #34of 51 Gemini 4 Argon82.4%
GDP.pdfWriting & content · Answer questions about real professional PDFs 16.0% #29of 38 GPT-6.1 Sol35.0%
SimpleQA VerifiedWriting & content · Factual accuracy on 1,000 short questions 60.3% #16of 66 Gemini 4 Argon76.0%
WebDev ArenaDesign & visual · People vote on which model builds the better website (Elo) 1534 #38of 88 Claude Opus 5.51820
SimpleBenchReasoning & knowledge · Common-sense trick questions about space, time and people 74.5% #7of 43 Claude Fable 581.9%
CritPtReasoning & knowledge · Open-ended, research-level physics problems 17.7% #32of 96 GPT-5.6 Sol32.3%
Chess PuzzlesReasoning & knowledge · Find the single best move in 100 new puzzles 27.0% #49of 70 GPT-6 Astra76.0%
LMCAReasoning & knowledge · Judge the quality of conceptual arguments like experts 48.4% #30of 96 Claude Sonnet 5.568.3%
ProofBenchMaths · Graduate maths proofs verified in Lean 4 43.0% #38of 70 Claude Sonnet 5.5100.0%

Specs

Details

Model ID
meta/muse-spark-1.2
Developer
Meta
Release date
Aug 5, 2026 ·
Context window
1,048,576 tokens
Max output
943,718 tokens
API price
$1.25 / $4.25 per 1M input / output tokens
Providers
1
Input types
text, image, video, file
Reasoning
Yes
Tool calling
Yes
Open weights
No
Free tier
No

Release date, prices and providers from OpenRouter's catalog; buying directly from a provider can cost a little differently. Uptime is the share of OpenRouter requests each provider served over the last day. Benchmark scores are the best result per benchmark from Epoch AI, licensed CC BY 4.0. Rank counts every model tracked here with a score.