Gemma 4 12B gemma-4-12b-it
google · gemma efficient
#52 overall
- params
- 12B
- arch
- dense
- license
- apache-2.0 open weights
- released
- May 2026
- downloads/30d
- 1.7M
# architecture
- layers
- 48
- d_model
- 3840
- heads
- 16q / 8kv
- head dim
- 256
- vocab
- 262144
- family
- gemma4_unified_text
signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.
# why ranked
Overall: #52 — score 49.9 ● high — all signals present
| signal | weight | input (0–100) |
|---|---|---|
| aa_intelligence | 0.70 | 51.2 |
| bench_composite | 0.30 | 46.9 |
benchmark panel evidence:
| complete panel | score (0–100) |
|---|---|
| aa-coding-v1 | 48.4 |
| aime-2026-v1 | 45.4 |
Coding: provisional — score 38.2 ● low — a single signal; treat with caution
provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.
| signal | weight | input (0–100) |
|---|---|---|
| aa_coding | 0.30 | 38.2 |
| aider_polyglot | 0.30 | — |
| swe_bench_verified | 0.40 | — |
missing signals are dropped and the remaining weights renormalized — never imputed.
# trend
methodology changed during this history window; score movement across that boundary is not model movement. See methodology v6.
# benchmarks
| benchmark | score | source | date |
|---|---|---|---|
| AA Coding Index | 31.0 | Artificial Analysis | — |
| AA Intelligence Index | 14.2 | Artificial Analysis | — |
| AIME 2026 | 0.8 / 1 | LLM Stats | — |
| Big-bench-extra-hard | 0.5 / 1 | LLM Stats | — |
| Codeforces | 0.6 / 3000 | LLM Stats | — |
| Covost2 | 0.4 / 1 | LLM Stats | — |
# where to run
| provider | quant | ctx | $/M in | $/M out | $/M cache | price src | tps | uptime | ✓ |
|---|---|---|---|---|---|---|---|---|---|
| SiliconFlow | — | — | — | — | siliconflow | — | — | ✓ | |
| Sail | 16k | $0.30 | $2.00 | — | via litellm | — | — |
sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price
source aliases
- aa
gemma-4-12b- hf
google/gemma-4-12B- openrouter
google/gemma-4-12b-it