Gemma 3n
Google's mobile-first open-weight multimodal model with audio, image, video and text input.
About
Gemma 3n is Google's on-device model family, released June 2025, offering multimodal input (text, image, audio and video) on phones, laptops and edge devices. It uses MatFormer architecture and Per-Layer Embedding caching so the E2B variant runs on about 2GB of memory with under 2B effective parameters. Strengths are offline operation, 140-language text support and broad tooling; honest limits are the 32K context window and that the original launch checkpoint only supported text and vision input, with full multimodal features rolling out after.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: Official Google AI docs read: 32K context, E2B/E4B effective params, PLE caching, MatFormer confirmed; Gemma family crossed 160M downloads; $150K Impact Challenge drove developer adoption; Gemma Terms of Use allows commercial use but is not an OSI-approved open license; Provider publishes no numeric benchmark table on the official docs page; press LMArena claims not provider-published. License: Open weights under Google's Gemma Terms of Use (responsible commercial use allowed, not OSI-open)..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 5.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Official Google AI docs read: 32K context, E2B/E4B effective params, PLE caching, MatFormer confirmed
- · Gemma family crossed 160M downloads; $150K Impact Challenge drove developer adoption
- · Gemma Terms of Use allows commercial use but is not an OSI-approved open license
- · Provider publishes no numeric benchmark table on the official docs page; press LMArena claims not provider-published