Benchmark Almanac

Meta

Best on the Capabilities Index Muse Wonder 161.7
39 models 17 model cards read ai.meta.com

Meta's models against the field

Every model on the Epoch Capabilities Index by release date, with Meta's in color.

50 75 100 125 150 175 2023 2024 2025 2026 2027 Muse Spark Open Muse Wonder Llama 4 Maverick 50 75 100 125 150 175 2023 2024 2025 2026 2027 Muse Spark Open Llama 4 Maverick

Its strongest models

Meta Oct 2026

Muse Wonder

Closed weights

161.7 #13 of 271
Epoch Capabilities Index
  • Agents' Last Exam 38.5% #7 /12
  • HealthBench Professional 60.5% #15 /25
  • Harvey LAB (Legal Agent Benchmark) 17.1% #7 /80
Meta Sep 2026

Muse Spark 1.3

Closed weights Flagship

157.4 #24 of 271
Epoch Capabilities Index
  • Terminal-Bench 4.0 14.6% #20 /23
  • AA-Omniscience 50.8% #4 /17
  • Harvey LAB (Legal Agent Benchmark) 23.8% #2 /80
Meta Aug 2026

Muse Spark 1.2

Closed weights Flagship

155.2 #38 of 271
Epoch Capabilities Index
  • DeepSWE 54.9% #17 /28
  • Harvey LAB (Legal Agent Benchmark) 25.4% #1 /80
  • ProgramBench 0% #24 /25
Meta Oct 2026

Muse Spark Open

Open weights

155.1 #39 of 271
Epoch Capabilities Index
  • SciCode 58.5% #5 /47
  • MultiChallenge 74.2% #6 /35
  • SimpleVQA 73.6% #5 /23
Meta Jul 2026

Muse Spark 1.1

Closed weights Flagship

154.2 #46 of 271
Epoch Capabilities Index
  • SWE-Bench Pro 61.5% #1 /24
  • DeepSWE 53.3% #21 /28
  • MultiChallenge 75.3% #4 /35
Meta Apr 2026

Muse Spark

Closed weights Flagship

152.2 #52 of 271
Epoch Capabilities Index
  • Humanity's Last Exam 40.6% #7 /47
  • SWE-Bench Pro 55% #3 /24
  • MMMU-Pro 80.4% #12 /72

What its model cards measure

Tall ticks are Meta model cards that list the benchmark in their headline table. Gray dots are cards that leave it out. Rows are the benchmarks the lab listed most, at least twice each, in the order it started using them.

MMLU BIG-Bench Hard AGIEval GPQA Diamond MATH HumanEval GSM8K ARC-Challenge MMLU-Pro MBPP MGSM IFEval MMMU-Pro MMMU SWE-bench Verified Humanity's Last Exam CharXiv DeepSearchQA SWE-Bench Pro DeepSWE JobBench Terminal-Bench 2.1 MCP Atlas OSWorld-Verified Vals Finance Agent v2 GDPval-AA v2 AutomationBench Artificial Analysis Intelligence Index v4.3

Model cards

DateModelDocumentBenchmarks
15 Oct 2026 Muse Spark Open Muse Spark Open model card (Hugging Face) and launch post 24
12 Oct 2026 Muse Wonder Muse Wonder: A Frontier Model For Personal & Professional AI Agents (Meta) 15
12 Oct 2026 Muse Wonder Full Muse Wonder: A Frontier Model For Personal & Professional AI Agents (Meta) 15
2 Sep 2026 Muse Spark 1.3 Introducing Muse Spark 1.3 9
10 Aug 2026 Muse Glimmer Introducing Muse Glimmer: An Open Agentic Model That Runs on Your Device (Meta) and model card 24
5 Aug 2026 Muse Spark 1.2 Introducing Muse Code and Muse Spark 1.2 2
9 Jul 2026 Muse Spark 1.1 Introducing Muse Spark 1.1 and the Meta Model API 11
8 Apr 2026 Muse Spark Introducing Muse Spark (Meta Superintelligence Labs) 19
24 Sep 2025 Code World Model (CWM) Code World Model (CWM) README and model card (GitHub, facebookresearch/cwm) 6
11 Jun 2025 V-JEPA 2 Introducing V-JEPA 2 world model and new benchmarks for physical reasoning (Meta AI blog) and GitHub README 2
5 Apr 2025 Llama 4 Scout Llama 4 Scout 17B-16E Instruct model card (Hugging Face) 14
5 Apr 2025 Llama 4 Maverick Llama 4 model card (Hugging Face meta-llama/Llama-4-Maverick-17B-128E-Instruct) 14
5 Apr 2025 Llama 4 Behemoth The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation 6
6 Dec 2024 Llama 3.3 70B Instruct Llama 3.3 model card (Hugging Face meta-llama/Llama-3.3-70B-Instruct) 9
23 Jul 2024 Llama 3.1-405B The Llama 3 Herd of Models (tech report), Table 2 15
18 Apr 2024 Llama 3-70B Introducing Meta Llama 3 (blog) + meta-llama/llama3 MODEL_CARD.md 15
18 Jul 2023 Llama 2-70B Llama 2 model card (Hugging Face meta-llama/Llama-2-70b-hf) 3

Every Meta model