Gemma 4 E4B
#8 v Small & Edge modelygoogle · v4 · e4b · od 2. April 2026 · 11× · naposledy 30. 6. 2026
30
Momentum
Gemma 4 E4B is an open-weight edge language model from Google DeepMind with approximately 4.5 billion effective parameters (8B total), designed for on-device deployment on mobile devices and laptops. It natively supports text, image, video, and audio input, and uses Per-Layer Embeddings (PLE) for maximum parameter efficiency on edge hardware. The model is available under the Apache 2.0 license and can run fully offline. It is part of the Gemma 4 family, which includes four sizes: E2B, E4B, 26B A4B, and 31B.
Vývoj momenta
04.04.03.07.
Vlastnosti
| Key Benchmark (%) | MMLU-Pro: 69.4% | AIME 2026: 42.5% | LiveCodeBench v6: 52.0% | MMMU Pro (Vision): 52.6% (each E4B-specific) |
| Context Window (Tokens) | 128,000 tokens |
| License | Apache 2.0 (unrestricted commercial use, fine-tuning, redistribution) |
| Multimodality | Text, image (variable resolution/aspect ratio), video (frame sequences), audio (ASR & speech-to-text translation) – all natively integrated in the model |
| Platform | On-device (Android, iOS, Desktop, IoT, Web) via LiteRT-LM; Hugging Face, Kaggle, Ollama, llama.cpp, LM Studio, vLLM, MLX, Unsloth, SGLang; Google AI Edge Gallery |
| Price | Free (open-weight model, weights freely downloadable) |
| Release Date | April 2, 2026 |