Models · Google (Gemini / DeepMind) · Out sinceReleased 3 Jun 2026
Gemma 4 12B IT
Gemma 4 12B IT is made by Google (Gemini / DeepMind). We don't have enough test results yet to rank it. Its makers have published it, so you can run it on your own servers.
Gemma 4 12B 'Unified': dense 11.95B, encoder-free (image patches and audio projected straight into the LLM), Apache 2.0, configurable thinking. Google says it runs in 16GB VRAM/unified memory. HF repo created 2026-05-23; public launch 2026-06-03. Pretrained base google/gemma-4-12B published. Not on OpenRouter; no Google list API price.
- —
- —
- —
- Price unknown— / —
- —
- 262K
- —
- 2 (2 independent2 indep.)
- 3 Jun 2026
- Google (Gemini / DeepMind)
- text, image, audio
Scores are out of 100 for each area, compared with every model we track; “#” is its rank. Context is how much text it can read at once (1M is roughly 700,000 words). Blended price mixes the cost of what you send and what it writes back.
Run it yourself
Self-hosting
On your own servers,Licence, sizeunder your own control.and hardware.
- Yes
- 11.95B
Runs on a powerful laptop or workstation.
≤ 24 GB at 4-bit (one consumer GPU / 32 GB Mac). Weights ≈ 13 GB at 8-bit, 7 GB at 4-bit (+10–30% for KV cache).
Languages: Languages: 35+ out of the box, pretrained on 140+. EuroEval Swedish: IT 1.81, base 2.58.
Training it further: Fine-tuning: Card: encoder-free design lets the entire model be fine-tuned in one pass; Google QLoRA guide; Unsloth.
Quantisations: bf16, qat q4_0 gguf (google/gemma-4-12B-it-qat-q4_0-gguf), qat w4a16 (google/gemma-4-12B-it-qat-w4a16-ct), gguf (community: unsloth). Engines: Transformers, llama.cpp.
Download (Hugging Face)Weights on Hugging Face ↗huggingface.co ↗blog.google ↗
Every test result
Every result
All the numbers,with where they came from.
Every result we've found for this model, with who measured it and a link to where we read it.
All raw rows (vendor and independent kept separate), with setting label, source, date, harness and notes.
| NotesNotes | ||||||
|---|---|---|---|---|---|---|
| Artificial Analysis Intelligence Index | 14.2 | DefaultGemma 4 12B (Reasoning) | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug gemma-4-12b; AA flags this Intelligence Index as ESTIMATED (intelligenceIndexIsEstimated=true, not all component evals run) - treat as provisional |
| EuroEval Swedish (generative) | 1.8 | Defaultgoogle/gemma-4-12B-it (val) | Independent testIndependentEuroEval (Alexandra Institute) ↗ | 29 Sep 2026 | — | rank tier 7; ±0.11; lower is better; SweDN 38.73, Skolprov 45.59, Swedish facts 32.94, ScaLA-sv 59.78 |