cd /news/large-language-models/muse-spark-1-3 · home › topics › large-language-models › article
[ARTICLE · art-173631] src=dev.meta.ai ↗ pub= topic=large-language-models verified=true sentiment=↑ positive

Muse Spark 1.3

Muse Spark 1.3, a model trained for long-horizon agentic workflows and competitive coding, was released with a 1M-token context window and benchmark results the company reports ahead of Muse Spark 1.2 on most agentic tasks. On the internal Agentic IF Index, Muse Spark 1.3 (max) scored 98.5 versus 66.3 for Muse Spark 1.2 (xhigh), and it posted 75.4 on DeepSWE v1.1 long-horizon agentic coding against 55.0 for 1.2, while GPT 5.6 Sol (max) scored 73.0 and Opus 5 (max) 74.0 on the same benchmark. The model is priced at $1.25 per million input tokens, $0.15 per million cached input tokens and $4.25 per million output tokens, with a $0.10/$0.002/$0.20 contributor tier whose data is used to improve the company's products.

key takeaway

Muse Spark 1.3 offers a 1M-token context window at $1.25 per million input and $4.25 per million output tokens, or $0.10/$0.20 on the contributor tier whose data is used for training, and scores 98.5 on the internal Agentic IF Index versus 66.3 for Muse Spark 1.2.

read1 min views5 publishedNov 21, 2026
Muse Spark 1.3
Image: source

Muse Spark 1.3 is trained for agentic workflows and optimized for competitive coding performance. Developers can expect higher first-attempt accuracy and reliable tool calling.

Meet Muse Spark 1.3 #

Trained for long-horizon, agentic workflows

Competitive coding performance

Native multimodal perception

Muse Spark

Introducing Muse Spark 1.3

Introducing Muse Spark 1.3, with max reasoning for challenging reasoning and agentic tasks and improved real-world usability.

Muse Spark 1.3 benchmarks #

Category

Benchmark

Muse Spark 1.3 (max) Meta

Muse Spark 1.2 (xhigh) Meta

GPT 5.6 Sol (max) OpenAI

Opus 5 (max) Anthropic

Agent

Long Context

Coding

GDPVal-AA v2Knowledge work

JobBenchProfessional tool use

OSWorld 2.0Agentic computer use

DeepSearchQAAgentic browsing

Agentic IF Index (Internal)Instruction following

AutomationBenchE2E business workflows

MRCR 256K–512KLong context

MRCR 512K–1MLong context

DeepSWE v1.1Long-horizon agentic coding

SWEAtlas CodeBase QnACodebase understanding

Terminal-Bench 2.1Agentic terminal coding

1754

64.9

66.9(partial)32.0(binary) 90.3

57.8

49.6

98.5

98.1

75.4

59.4

88.8

1615

61.6

47.6(partial)17.9(binary) 85.9

46.2

38.2

66.3

55.5

55.0

46.2

82.9

1710

45.4

62.7(partial)27.3(binary) 93.1

60.5

46.7

91.5

73.8

73.0

53.5

88.8

1824

65.7

68.3(partial)31.4(binary) 90.4

59.1

50.3

74.0

52.7

86.7

GET STARTED

More ways to build

Models and pricing #

Model

Context window

Input (Mtok)

Cached input (Mtok)

Output (Mtok)

muse-spark-1.3-contributorUsed to improve our products.

1M

$0.10

$0.002

$0.20

muse-spark-1.3Not used to improve our products.

1M

$1.25

$0.15

$4.25

── more in #large-language-models 4 stories · sorted by recency
── more on @muse spark 1.3 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/muse-spark-1-3] indexed:0 read:1min 2026-11-21 · —