Organization

Meta AI

Tracked models

101

Frontier models

2

Largest training run

5.2 x 10^25 FLOP

101 tracked AI models list Meta AI as a developer. 2 of them are frontier models — among the largest training runs of their moment. The largest is Llama 4 Behemoth (preview), at 5.2 x 10^25 FLOP (April 2025). The most recent tracked release is from July 2026.

It operates 8 tracked GPU clusters.

Models (80+)

Llama 4 Behemoth (preview)5.2 x 10^25 FLOP Llama 3.1-405B3.8 x 10^25 FLOP Llama 3.1-70B7.9 x 10^24 FLOP Llama-3.1-Nemotron-70B-Instruct7.9 x 10^24 FLOP Llama 3-70B7.9 x 10^24 FLOP Llama 3.3 70B6.9 x 10^24 FLOP Llama 4 Scout4.1 x 10^24 FLOP Llama 4 Maverick2.2 x 10^24 FLOP Movie Gen Video1.6 x 10^24 FLOP Code Llama-70B1.3 x 10^24 FLOP Llama 3.1-8B1.2 x 10^24 FLOP Llama 2-70B8.1 x 10^23 FLOP Llama 3-8B7.2 x 10^23 FLOP Llama 3.2 11B5.8 x 10^23 FLOP LIMA5.5 x 10^23 FLOP LLaMA-65B5.5 x 10^23 FLOP Code Llama-34B5.3 x 10^23 FLOP OPT-IML (175B)4.3 x 10^23 FLOP BlenderBot 34.3 x 10^23 FLOP OPT-175B4.3 x 10^23 FLOP Llama 2-34B4.1 x 10^23 FLOP OPT-13B3.5 x 10^23 FLOP Galactica3.2 x 10^23 FLOP LLaMA-33B2.7 x 10^23 FLOP SAM 32 x 10^23 FLOP Llama 3.2 3B1.7 x 10^23 FLOP Llama Guard1.6 x 10^23 FLOP Llama 2-13B1.6 x 10^23 FLOP Movie Gen Audio1.4 x 10^23 FLOP OPT-66B1.1 x 10^23 FLOP Code Llama-7B1.1 x 10^23 FLOP Llama 2-7B8.4 x 10^22 FLOP LLaMA-13B7.8 x 10^22 FLOP ESM2-15B7.4 x 10^22 FLOP Llama 3.2 1B6.6 x 10^22 FLOP OPT-30B5.4 x 10^22 FLOP LLaMA-7B4 x 10^22 FLOP Fairseq-dense 13B3.3 x 10^22 FLOP ESM2-3B3 x 10^22 FLOP XGLM-7.5B2.3 x 10^22 FLOP MoE-1.1T2.2 x 10^22 FLOP NLLB1.8 x 10^22 FLOP Segment Anything Model 21.2 x 10^22 FLOP OPT-6.7B1.2 x 10^22 FLOP AudioGen9.5 x 10^21 FLOP Segment Anything Model7.8 x 10^21 FLOP ESM2-650M7.6 x 10^21 FLOP Make-A-Scene6.4 x 10^21 FLOP ViT-G (model soup)3.4 x 10^21 FLOP V-JEPA1.6 x 10^21 FLOP ESM2-150M1.1 x 10^21 FLOP Sparse all-MLP5.3 x 10^20 FLOP DLRM-20213 x 10^20 FLOP ESM2-35M2.1 x 10^20 FLOP Contriever1.6 x 10^20 FLOP DeiT-B7.9 x 10^19 FLOP ConvS2S (ensemble of 8 models)5.6 x 10^19 FLOP ESM2-8M4.8 x 10^19 FLOP DiffQ Transformer (16L)3 x 10^19 FLOP MultiBand Diffusion2.6 x 10^19 FLOP Detic2.3 x 10^19 FLOP NoPos2.1 x 10^19 FLOP MemSizer (language modeling)7.3 x 10^18 FLOP Brain2Qwerty1.6 x 10^18 FLOP Muse Spark 1.1July 2026 Muse ImageJuly 2026 Muse SparkApril 2026 Meta's Generative Ads Model (GEM)November 2025 Llama 4 Scout + ScaleRLOctober 2025 DINOv3August 2025 Apollo 7BDecember 2024 Apollo 3BDecember 2024 Apollo 1.5BDecember 2024 LongVUOctober 2024 Llama 3.2 90BSeptember 2024 LLaMA-3-Instruct-8BApril 2024 MAGNeTJanuary 2024 Emu (Meta)September 2023 Code Llama-13BAugust 2023 MusicGenJune 2023

GPU clusters (8)

Meta 100k100K H100e Meta GenAI 2024b24.6K H100e Meta GenAI 2024a24.6K H100e Meta Research SuperCluster (RSC-1) Phase 25K H100e Meta Research SuperCluster 2 (RSC-2)2.5K H100e Meta Research SuperCluster (RSC-1) Phase 11.9K H100e Meta 2017 V100 Cluster1.4K H100e Meta 2017 P100 Cluster9.37 H100e

SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.