Model datasheet · Gemma 4
Gemma 4 E2B
The smallest Gemma 4, at 3.35GB in Google's QAT build — genuinely multimodal on hardware that cannot run anything else in this catalog. Use it for dictation, captioning and simple chat; its own vendor numbers rule out reasoning and tools.
- Vendor
- Architecture
- Dense · 5.1B
- Context
- 131,072 tokens
- License
- Apache-2.0
- Released
- 2026-07-02
- Vision
- Yes
→ GGUF (llama.cpp / LM Studio / Ollama) → MLX (Apple Silicon)
§1 Characteristics by quantization band
— Not yet rated · contributions welcome —
Q4–Q5 — 4–5 bit
3.2GB weights · e.g. Q4_0 , Q4_K_M , mlx-4bit + KV: 70MB at 8K · 238MB at 32K · 910MB at 128Kdata checked Aug 2026
- quant-degradation-community — community consensus @ Q4_0 · aggregated · 2026-08-30
Editorial derivation: QAT Q4_0 (3.35GB) is the intended build; sub-3B-effective models lose more per bit than large ones.
data checked Aug 2026
- quant-degradation-community — community consensus @ Q4_0 · aggregated · 2026-08-30
Editorial derivation: carried down one point from the full band.
data checked Aug 2026
- quant-degradation-community — community consensus @ Q4_0 · aggregated · 2026-08-30
Editorial derivation: do not build a tool-using agent on this.
— Not yet rated · contributions welcome —
FP16 — full precision (fp16/bf16)
10GB weights · e.g. bf16 + KV: 70MB at 8K · 238MB at 32K · 910MB at 128Kdata checked Aug 2026
- vendor-model-card — AIME 2026 (no tools) = 37.5 @ bf16 · vendor-reported · 2026-08-30
MMLU-Pro: 60.0, GPQA Diamond: 43.4 — the knowledge floor of this catalog.
data checked Aug 2026
- vendor-model-card — LiveCodeBench v6 = 44 @ bf16 · vendor-reported · 2026-08-30
Snippets and single functions only.
data checked Aug 2026
- vendor-model-card — Tau2 average (with tools) = 24.5 @ bf16 · vendor-reported · 2026-08-30
The vendor's own number says tool use does not work reliably at this size. Documented as a known issue rather than dressed up.
data checked Aug 2026
- vendor-model-card — MRCR v2 @128K (8 needle) = 19.1 @ bf16 · vendor-reported · 2026-08-30
Effectively no long-context retrieval.
§2 Known issues & what fixes them
Tau2 24.5 — the vendor's own number says roughly three in four tool-use episodes fail. Treat tool calling as unsupported on this model rather than as something to tune.
No prompt technique recovers this. If you need tools on small hardware, Gemma 4 E4B (42.2) is the minimum and the 12B (69.0) is the first size that actually works.
Evidence · 1 source · anecdotal
- Gemma 4 model card — Tau2 per variant (E2B 24.5) (vendor, 2026-08-30)
MMLU-Pro 60.0 and GPQA Diamond 43.4 are the lowest in this catalog. At 2.3B effective parameters there is simply not enough capacity to store reliable world knowledge — it will answer general-knowledge questions fluently and wrongly.
Give it the source text and let it summarize or extract. It is a competent reader; it is not a reference work.
Arithmetic belongs in a calculator, not in the weights — though note the model's own tool-calling is unreliable, so drive the tool from your harness rather than letting the model choose it.
Evidence · 1 source · anecdotal
- Gemma 4 model card — MMLU-Pro and GPQA Diamond per variant (vendor, 2026-08-30)
MRCR v2 at 128K is 19.1% — the advertised window is a buffer size, not a working memory.
Retrieve a page at a time.
Evidence · 1 source · anecdotal
- Gemma 4 model card — MRCR v2 @128K per variant (vendor, 2026-08-30)