Country

France

Tracked models

162

Frontier models

1

Largest training run

2.3 x 10^24 FLOP

162 tracked AI models have at least one contributing organization based in France. 1 of them is frontier models — among the largest training runs of their moment. The largest is Mixtral 8x22B, at 2.3 x 10^24 FLOP (April 2024). The most recent tracked release is from April 2026.

It hosts 14 tracked GPU clusters.

Models (80+)

Mixtral 8x22B2.3 x 10^24 FLOP Chameleon-34B1.6 x 10^24 FLOP Mistral Small 31.2 x 10^24 FLOP Mixtral 8x7B7.7 x 10^23 FLOP Multi-Token Prediction 7B3.8 x 10^23 FLOP BLOOM-176B3.7 x 10^23 FLOP FragLlama: Next-fragment prediction for molecular design3.3 x 10^23 FLOP LightOn Mini2.4 x 10^23 FLOP Multi-Token Prediction 13B1.5 x 10^23 FLOP XLMR-XXL3.4 x 10^22 FLOP Pleias-RAG-1B3 x 10^22 FLOP Pleias 1.0 1.2B3 x 10^22 FLOP XGLM-7.5B2.3 x 10^22 FLOP SEER1.8 x 10^22 FLOP BLOOM-7.1B1.5 x 10^22 FLOP Generative BST1.4 x 10^22 FLOP V-JEPA 29.1 x 10^21 FLOP DINOv27.4 x 10^21 FLOP HuBERT5.5 x 10^21 FLOP MSA Transformer5.5 x 10^21 FLOP ESM1b5.1 x 10^21 FLOP Reason-ModernColBERT3.7 x 10^21 FLOP GTE-ModernColBERT-v13.7 x 10^21 FLOP Incoder-6.7B3 x 10^21 FLOP Pleias-RAG-350m2.7 x 10^21 FLOP Pleias 1.0 350m2.7 x 10^21 FLOP KEPLER1.7 x 10^21 FLOP MACE-MP-08.8 x 10^20 FLOP Rita-XLarge8.6 x 10^20 FLOP CamemBERT8.3 x 10^20 FLOP ALiBi (L=3072, Lvalid = 3072)8.1 x 10^20 FLOP DD-PPO7.8 x 10^20 FLOP GemNet-OC5.4 x 10^20 FLOP ESM1-670M (UR50/D)4.8 x 10^20 FLOP Big Transformer for Back-Translation4.8 x 10^20 FLOP Masked Autoencoders ViT-H4.6 x 10^20 FLOP ESM1-670M (UR50/S)4.4 x 10^20 FLOP Transformer-XL + SIS3.8 x 10^20 FLOP PAGnol-XL2.6 x 10^20 FLOP ViT + DINO2.1 x 10^20 FLOP Contriever1.6 x 10^20 FLOP ESM1-670M (UR100)1.4 x 10^20 FLOP ESM1v1.4 x 10^20 FLOP PeTriBERT10^20 FLOP DeiT-B7.9 x 10^19 FLOP ESM1-85M5.6 x 10^19 FLOP Transformer (Adaptive Input Embeddings) WT1034.5 x 10^19 FLOP FAIRSEQ Adaptive Inputs3.2 x 10^19 FLOP Base LM + kNN LM + Continuous Cache3.1 x 10^19 FLOP ESM1-43M2.8 x 10^19 FLOP MultiBand Diffusion2.6 x 10^19 FLOP Sandwich Transformer2.4 x 10^19 FLOP All-attention network + adaptive span1.2 x 10^19 FLOP code2seq1.2 x 10^19 FLOP Novae1.1 x 10^19 FLOP MGK 4 heads (medium)8.9 x 10^18 FLOP Feedback Transformer7.7 x 10^18 FLOP DeLighT3.8 x 10^18 FLOP Shortformer3 x 10^18 FLOP MGK 8 heads (small)3 x 10^18 FLOP RetinaNet-R1012.1 x 10^18 FLOP Brain2Qwerty1.6 x 10^18 FLOP TransE1.3 x 10^18 FLOP DDPM9 x 10^17 FLOP DDIM9 x 10^17 FLOP AlphaX-18.9 x 10^17 FLOP VDCNN (on Amazon Review Full dataset)5.7 x 10^17 FLOP code2vec3.2 x 10^17 FLOP Pluribus6.6 x 10^16 FLOP Fisher Vector image classifier9.1 x 10^13 FLOP Mistral Medium 3.5April 2026 Devstral 2 (123B)December 2025 Devstral 2 (24B)December 2025 Mistral 3 LargeDecember 2025 Ministral 3 14BDecember 2025 Ministral 3 8BDecember 2025 Ministral 3 3BDecember 2025 Magistral Small 1.2September 2025 Magistral Medium 1.2September 2025 Mistral Medium 3.1August 2025

GPU clusters (14)

Sesterce Synapse Phase 24.1K H100e Sesterce H100s Phase 24.1K H100e CEA EXA1-HE Phase 32.7K H100e Sesterce Synapse Phase 12K H100e CEA EXA1-HE Phase 21.9K H100e Jean Zay Supercomputer Phase 41.6K H100e Scaleway Nabuchodonosor1K H100e GENCI Adastra262 H100e TotalEnergies Pangea III214 H100e Jean Zay Supercomputer Phase 2170 H100e Jean Zay Supercomputer Phase 3131 H100e Jean Zay Supercomputer Phase 165.94 H100e Atos Spartan260.54 H100e Commissariat a l'Energie Atomique Topaze60.54 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.