Ao vivo
Organization

Google

Tracked models
147
Frontier models
20
Largest training run
7.3 x 10^24 FLOP
Data-center compute
3.2M H100e

147 tracked AI models list Google as a developer. 20 of them are frontier models — among the largest training runs of their moment. The largest is PaLM 2, at 7.3 x 10^24 FLOP (May 2023). The most recent tracked release is from March 2026.

It owns 20 tracked AI data centers, holding about 3.2M H100e and drawing 2.71 GW. It operates 10 tracked GPU clusters.

It employs around 6,000 people.

Models (80+)
PaLM 27.3 x 10^24 FLOPMinerva (540B)2.7 x 10^24 FLOPFlan-PaLM 540B2.5 x 10^24 FLOPU-PaLM (540B)2.5 x 10^24 FLOPBIG-G 137B5.6 x 10^23 FLOPGemma 3n5.2 x 10^23 FLOPGLaM3.6 x 10^23 FLOPLaMDA3.5 x 10^23 FLOPGemma 1.1 7B Instruct3.1 x 10^23 FLOPST-MoE2.9 x 10^23 FLOPViT-22B1.9 x 10^23 FLOPPaLI1.7 x 10^23 FLOPSwitch8.2 x 10^22 FLOPmT5-XXL8.2 x 10^22 FLOPByT5-XXL8.1 x 10^22 FLOPProtT5-XXL7.4 x 10^22 FLOPGShard (dense)4.8 x 10^22 FLOPCoAtNet4.3 x 10^22 FLOPBASIC-L4.1 x 10^22 FLOPPaliGemma 2 3B Mix 2244 x 10^22 FLOPProtBERT-BFD3.9 x 10^22 FLOPProtT5-XXL-BFD3.7 x 10^22 FLOPFlan-T5 11B3.3 x 10^22 FLOPT5-11B3.3 x 10^22 FLOPNoisy Student (L2)2.6 x 10^22 FLOPProtT5-XL-U501.9 x 10^22 FLOPLIMoE-H/141.8 x 10^22 FLOPGShard (600B)1.3 x 10^22 FLOPSILC-S* (86M)10^22 FLOPT5-3B9 x 10^21 FLOPAlphaEarth Foundations (AEF)8.5 x 10^21 FLOPContextNet + Noisy Student8.2 x 10^21 FLOPConformer + Wav2vec 2.0 + Noisy Student7.6 x 10^21 FLOPProtBERT-UniRef7.3 x 10^21 FLOPGNMT6.6 x 10^21 FLOPViT-G (model soup)3.4 x 10^21 FLOPELECTRA3.1 x 10^21 FLOPALBERT-xxlarge2.4 x 10^21 FLOPAeneas2.3 x 10^21 FLOPMnasNet-A1 + SSDLite1.5 x 10^21 FLOPMnasNet-A31.5 x 10^21 FLOPTOME10^21 FLOPXception4.4 x 10^20 FLOPSNM-skip3 x 10^20 FLOPBERT-Large2.9 x 10^20 FLOPInception v310^20 FLOPEfficientNetV2-XL9.6 x 10^19 FLOPSeq2Seq LSTM5.6 x 10^19 FLOPFFN SwiGLU3.4 x 10^19 FLOPTAPE Transformer3 x 10^19 FLOPTransformer + Simple Recurrent Unit1.1 x 10^19 FLOPMetNet9.5 x 10^18 FLOPDistBelief NNLM2.6 x 10^18 FLOPGoogLeNet / InceptionV11.5 x 10^18 FLOPSB-LM1.4 x 10^18 FLOPTransE1.3 x 10^18 FLOPConditional Maximum Entropy Model (Gigaworld)10^18 FLOPKN-LM7.7 x 10^17 FLOPUnsupervised High-level Feature Learner6 x 10^17 FLOPCT-MoS + DynamicEval (WT2)5.4 x 10^17 FLOPCT-MoS (WT2)5.4 x 10^17 FLOPDistBelief Speech3.1 x 10^17 FLOPReLU-Speech1.3 x 10^17 FLOPWord2Vec (large)3.9 x 10^16 FLOPSearch-Proven Best LSTM3.3 x 10^15 FLOPGemini 3.1 Flash-LiteMarch 2026Gemini 2.5 Computer UseOctober 2025Gemini 2.5 Flash Image (Nano Banana)August 2025imagen 4 fastAugust 2025Gemini 2.5 Deep ThinkAugust 2025MedSigLIPJuly 2025SignGemmaMay 2025MedGemma 27BMay 2025Imagen 4May 2025Imagen 4 ultraMay 2025Sec-Gemini v1April 2025Chirp 3 Speech-to-TextMarch 2025Chirp 3 HD Text-to-SpeechMarch 2025PaliGemma 2 3B Mix 448February 2025Gemini 2.0 Flash ThinkingDecember 2024
Data centers (20)
GPU clusters (10)
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.
Other organizations