MDL
Gemma 4 12B
Verified against Artificial Analysis · Jun 15, 2026
Capability
74
percentile index
Price /M
$0.1 / $0.3
in / out
Context
—
Avg score
44.5
7 benchmarks
AA Intelligence
29.2
index
Speed
161 tok/s
TTFT 1.4s
Released
Jun 3, 2026
About
Gemma 4 12B is a open-weights model in the Gemma family from Google. Benchmarked on 7 evals, averaging 44.5.
Benchmark scores · 7
avg 44.5 — higher is betterThe research behind this model
all papers mentioning it →The Chain Holds, the Answer Folds: Trace-Answer Dissociation in Reasoning Models Under Adversarial Pressure▲ 1 on HF · May 27, 2026BioRefusalAudit: Auditing Biosecurity Refusal Depth Using General and Domain-Fine-Tuned Sparse AutoencodersMay 28, 2026Architecture-Sensitive Supervised Fine-Tuning for Screen-Conditioned Action Prediction: A PiSAR BenchmarkMay 28, 2026PrionNER: A Named Entity Recognition Dataset for Prion Disease Biomedical LiteratureMay 27, 2026ConRAG: Consensus-Driven Multi-View Retrieval for Multi-Hop Question AnsweringMay 27, 2026
Mentions matched by name in title/abstract — from the arXiv + HF daily corpus.
Performance & price
Output speed161 tok/s
Latency (TTFT)1.4s
First answer14s
Price in / out$0.1 / $0.3 per 1M
Blended price$0.15 / 1M
Coding index24.9
Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →
History
full ledger →Capability over time
Humanity's Last Exam14.6 → 14.8
7d ago
τ²-bench34.8 → 36.3
7d ago
Append-only ledger — every observed change to this model's numbers.
API providers
Artificial Analysis$0.3/M