{"id":"google-gemma-4-e4b-it","nom":"Gemma 4 E4B","editeur":"Google","pays":"États-Unis","poids_ouverts":true,"poids_selon":null,"licence":"Apache 2.0","licence_selon":null,"liens":{"page":"https://huggingface.co/google/gemma-4-E4B-it"},"fiche_auto":false,"page":"https://quelleia.com/voix/modeles/google-gemma-4-e4b-it/","releve":"2026-10-19","en_bref":"36e sur 39 en transcription en français ; 51e sur 53 en transcription en anglais (Open ASR).","presentation":"Modèle ouvert de la famille Gemma 4 de Google DeepMind, en version ajustée aux instructions, conçu pour les téléphones et appareils locaux, qui accepte texte, image et audio en entrée. Sur l'audio, il fait de la reconnaissance vocale et de la traduction de la parole vers du texte.","sortie":"2026-04-02","taille":"4,5 milliards de paramètres effectifs (8 milliards avec les tables d'embeddings)","langues":{"annoncees":{"annoncees":"plus de 35 langues prises en charge d'emblée (préentraînement sur plus de 140)","nombre":35,"plus":true},"prises_en_charge_selon_open_tts":null,"evaluees":["fr","en"]},"preuves":[{"appuie":["presentation"],"lu":"https://huggingface.co/google/gemma-4-E4B-it/raw/main/README.md (« Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. »), 17/10/2026"},{"appuie":["presentation"],"lu":"https://huggingface.co/google/gemma-4-E4B-it/raw/main/README.md (« **Audio** (E2B, E4B, and 12B only) - Automatic speech recognition (ASR) and speech-to-translated-text translation across multiple languages. »), 17/10/2026"},{"appuie":["presentation"],"lu":"https://huggingface.co/google/gemma-4-E4B-it/raw/main/README.md (« targeting deployment scenarios from mobile and edge devices (E2B, E4B) to consumer GPUs and workstations (12B, 26B A4B, 31B) »), 17/10/2026"},{"appuie":["taille"],"lu":"https://huggingface.co/google/gemma-4-E4B-it/raw/main/README.md (« | **Total Parameters** | 2.3B effective (5.1B with embeddings) | 4.5B effective (8B with embeddings) | »), 17/10/2026"},{"appuie":["langues"],"lu":"https://huggingface.co/google/gemma-4-E4B-it/raw/main/README.md (« **Multilingual** - Out-of-the-box support for 35+ languages, pre-trained on 140+ languages. »), 17/10/2026"},{"appuie":["sortie"],"lu":"https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/ (« Gemma 4: Byte for byte, the most capable open models Apr 02, 2026 »), 17/10/2026"},{"appuie":["presentation","sortie"],"lu":"https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/ (« Additionally, the E2B and E4B models feature native audio input for speech recognition and understanding. »), 17/10/2026"}],"classements":[{"id":"transcription/francais","nom":"Transcription en français","rang":36,"sur":39,"valeur":10.31,"unite":"%","chiffres":{"rtfx":133.83,"jeux":{"Common Voice":13.99,"FLEURS":7.42,"MLS":9.53}},"voisins":["meta-omniasr-ctc-1b-v2","openmoss-team-moss-transcribe-diarize"],"source":{"nom":"Open ASR Leaderboard","mention":"Source : Open ASR Leaderboard, Hugging Face","url":"https://huggingface.co/spaces/hf-audio/open_asr_leaderboard","releve":"2026-10-19"}},{"id":"transcription/open-asr-anglais","nom":"Transcription en anglais, Open ASR","rang":51,"sur":53,"valeur":11.86,"unite":"%","chiffres":{"rtfx":161.39,"jeux":{"AMI-Cleaned":20.57,"Earnings22-Cleaned-AA-chunked":14.16,"Gigaspeech-Cleaned":11.74,"LS Clean":3.85,"LS Other":9.97,"Private (conversational)":19.17,"Private (scripted)":8.88,"SPGISpeech":5.49,"URGENT24":26.38,"Voice Arena Monsoon":6.19,"Voxpopuli-AA-Cleaned":4.09}},"voisins":["meta-omniasr-ctc-7b-v2","abr-ai-niagara-19m-batch-en"],"source":{"nom":"Open ASR Leaderboard","mention":"Source : Open ASR Leaderboard, Hugging Face","url":"https://huggingface.co/spaces/hf-audio/open_asr_leaderboard","releve":"2026-10-19"}}],"noms_dans_les_sources":{"hf_audio":"google/gemma-4-E4B-it"}}