MDL
Gemma 3 4B Instruct
Verified against Artificial Analysis · Jul 13, 2026
Capability
30
percentile index
Price /M
$0 / $0
in / out
Context
—
Avg score
20.3
11 benchmarks
AA Intelligence
1.1
index
Weights
Open
Released
Mar 12, 2025
About
Gemma 3 4B Instruct is a open-weights model in the Gemma family from Google. Benchmarked on 11 evals, averaging 20.3.
Benchmark scores · 11
avg 20.3 — higher is betterThe research behind this model
all papers mentioning it →Logit-Contribution Scoring Identifies Non-Literal Retrieval Heads▲ 18 on HF · Jul 1, 2026Prompt-Level Distillation: A Non-Parametric Alternative to Model Fine-Tuning for Efficient Reasoning▲ 9 on HF · Jun 2, 2026Bag of Dims: Training-Free Mechanistic Interpretability via Dimension-Level Sign Patterns▲ 2 on HF · Jun 17, 2026Pressure-Testing Deception Probes in LLMs: Scaling, Robustness, and the Geometry of Deceptive Representations▲ 2 on HF · May 27, 2026Internal Representation, Not Clinical Knowledge: Where Apparent LLM Triage Failures OriginateMay 28, 2026
Mentions matched by name in title/abstract — from the arXiv + HF daily corpus.
Performance & price
Price in / out$0 / $0 per 1M
Coding index2.7
Math index12.7
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
Append-only ledger — every observed change to this model's numbers.
API providers
Artificial Analysis$0/M