# Gemma 4 E2B ![Organization](https://img.shields.io/badge/Org-Google-4285F4) ![Effective Params](https://img.shields.io/badge/Effective-2B-1f6feb) ![License](https://img.shields.io/badge/License-Gemma%20ToU-d29922) ![Weights](https://img.shields.io/badge/Weights-Open-2da44e) ![On--Device](https://img.shields.io/badge/Target-On--Device-2da44e) ![Released](https://img.shields.io/badge/Released-Dec%202025-555) > **Quick answer:** Gemma 4 E2B is Google's December 2025 smallest on-device multimodal model with 2B effective parameters, scoring 60% MMLU-Pro. It succeeds Gemma 3n E2B (40.5% MMLU-Pro) with a significant +19.5pp improvement at the same parameter scale. ## At a Glance **Where Gemma 4 E2B leads** - 60% MMLU-Pro — major improvement over Gemma 3n E2B (40.5%) - Smallest on-device Gemma 4 model - Multimodal capability at 2B effective params - September 2025 knowledge cutoff **Where it lags** - 60% MMLU-Pro — below Gemma 4 E4B (69.4%) at this generation - 2B scale limits knowledge and reasoning - Gemma Terms of Use **Best for:** Most resource-constrained on-device deployments; applications where E4B doesn't fit. ## What Gemma 4 E2B Is Gemma 4 E2B is Google's smallest model in the Gemma 4 family, targeting the most constrained on-device deployment scenarios. At 2B effective parameters, it achieves 60% MMLU-Pro — a substantial +19.5pp improvement over the predecessor Gemma 3n E2B (40.5%). For most on-device applications, Gemma 4 E4B (69.4% MMLU-Pro) provides meaningfully better performance at comparable device footprint. E2B is reserved for devices where E4B is too large. ## Specifications | Field | Value | |---|---| | Organization | Google | | Effective parameters | 2B | | License | Gemma Terms of Use | | HuggingFace | google/gemma-4-E2B-it | | Release date | December 2025 | | Knowledge cutoff | September 2025 | | Modality | Multimodal | ## Pricing Open weights under Gemma Terms of Use — free to self-host. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [MMLU-Pro](/benchmarks/mmlu-pro) | 60% | Benchgen evaluation | 2025-12 | ## Gemma 4 E2B vs Alternatives | Model | MMLU-Pro | Effective Params | License | |---|---|---|---| | **Gemma 4 E2B** | 60% | 2B | Gemma ToU | | [Gemma 3n E2B](/models/google/gemma-3n-e2b) | 40.5% | 2B | Gemma ToU | | [Gemma 4 E4B](/models/google/gemma-4-e4b) | 69.4% | 4B | Gemma ToU | Gemma 4 E2B vs Gemma 3n E2B: +19.5pp MMLU-Pro. vs Gemma 4 E4B: lower MMLU-Pro (60% vs 69.4%) but smaller footprint. ## Related Models & Benchmarks - Compare: [Gemma 3n E2B](/models/google/gemma-3n-e2b) · [Gemma 4 E4B](/models/google/gemma-4-e4b) · [Gemma 4 26B A4B](/models/google/gemma-4-26b-a4b) - Benchmarks: [MMLU-Pro](/benchmarks/mmlu-pro) ## Frequently Asked Questions

What is Gemma 4 E2B?

Google's December 2025 smallest on-device multimodal model with 2B effective parameters scoring 60% MMLU-Pro — a +19.5pp improvement over the predecessor Gemma 3n E2B.

Should I use Gemma 4 E2B or E4B?

Gemma 4 E4B for better performance (69.4% vs 60% MMLU-Pro). Use E2B only when E4B won't fit on the target device.

--- *Specs from Google's Gemma 4 release (December 2025) and Benchgen evaluations. Last updated 2026-07-24.*

Gemma 4 E2B — Training jobs

Training jobs