Gemma 4 31B Dense GGUF
Certified GGUF precision quantization ladder for google/gemma-4-31b dense architecture.
Certified GGUF precision variants (Q8_0, Q4_K_M, Q2_K) of google/gemma-4-31b evaluated across standard reasoning, mathematics, and science benchmarks (ARC-MC, GSM8K, MATH-500) under reproducible seed controls.
Architecture
- Parameters
- 30.70B
- Active
- 30.70B (100%)
- Layers
- 60
- Quantization ladder
- Certified GGUF
Downloads
Exact release files
3 immutable artifacts
gemma-4-31b.q4_k_m.gguf
Size: 19 GB
20–24 GB reference capacity (single RTX 3090/4090). Balanced performance and retention.
SHA-256 Checksum
7584d5b098322d7d7ebe0aae2a962c3371c88a547f6dc9fe475c61bd92b5e316
gemma-4-31b.q8_0.gguf
Size: 33 GB
32–48 GB reference capacity. Near-lossless precision for production and critical reasoning.
SHA-256 Checksum
22a8ab2d7a3d4b40ce0709e868c6d8953c51b815dee545c72d74ce3900075bc3
gemma-4-31b.q2_k.gguf
Size: 12 GB
12–16 GB reference capacity. Memory-constrained consumer hardware.
SHA-256 Checksum
e8e9bcfe6e1a57f7cf098f0922684ac7794257f71d4acc5319892fe9b811d9b6
Deployment
Quickstart & CLI
llama.cpp & Ollama
Stream weights directly from Hugging Face via Ollama CLI:
ollama run hf.co/Omega-Factory/gemma-4-31b-GGUF:Q4_K_MRun inference locally using llama.cpp with the recommended artifact:
llama-cli -hf Omega-Factory/gemma-4-31b-GGUF -m gemma-4-31b.q4_k_m.gguf -p "Explain the core mechanics of mixture-of-experts models." -n 512Launch a local OpenAI-compatible HTTP server:
llama-server -hf Omega-Factory/gemma-4-31b-GGUF -m gemma-4-31b.q4_k_m.gguf --port 8080 -ngl 99Evidence
Measured outcomes & precision ladder
Precision & Accuracy Matrix vs. FP16 Baseline
Holdout evaluation split (N=200 per category, seed 42) comparing all quant variants to unquantized baseline.
ARC-MC
85.0%
GSM8K
86.0%
MATH-500
36.5%
ARC-MC
95.0%
+10.0 pp
GSM8K
84.5%
-1.5 pp
MATH-500
37.0%
+0.5 pp
ARC-MC
94.5%
+9.5 pp
GSM8K
87.5%
+1.5 pp
MATH-500
35.5%
-1.0 pp
ARC-MC
91.0%
+6.0 pp
GSM8K
2.0%
-84.0 pp
MATH-500
6.0%
-30.5 pp
Item-Level McNemar Discordance & Flips vs. FP16 Baseline
Per-item analysis measuring individual test questions that flipped answer status relative to the unquantized FP16 baseline (N=200).
Recommended Release Rung Scoped Verification (Q4_K_M)
Formal cryptographic release marker evaluation records with 95% confidence intervals and sample counts.
ARC-Challenge Science Reasoning
gemma-4-31b-arc-mc-v1 · accuracy_pct
Base
85.00%
Release
94.50%
Delta
+9.50 pp
Holdout evaluation set (N=200, seed 42) comparing Q4_K_M quantized release against uncompressed FP16 baseline under standard multiple-choice log-likelihood scoring.
95% CI [3.63, 15.37] pp · 200 samples · 1 seeds · artifact 7584d5b098322d7d7ebe0aae2a962c3371c88a547f6dc9fe475c61bd92b5e316
GSM8K Grade School Math
gemma-4-31b-gsm8k-v1 · accuracy_pct
Base
86.00%
Release
87.50%
Delta
+1.50 pp
Holdout evaluation set (N=200, seed 42) comparing Q4_K_M quantized release against uncompressed FP16 baseline under greedy decoding.
95% CI [-5.14, 8.14] pp · 200 samples · 1 seeds · artifact 7584d5b098322d7d7ebe0aae2a962c3371c88a547f6dc9fe475c61bd92b5e316
MATH-500 Advanced Mathematics
gemma-4-31b-math500-v1 · accuracy_pct
Base
36.50%
Release
35.50%
Delta
-1.00 pp
Holdout evaluation set (N=200, seed 42) comparing Q4_K_M quantized release against uncompressed FP16 baseline under greedy decoding.
95% CI [-10.41, 8.41] pp · 200 samples · 1 seeds · artifact 7584d5b098322d7d7ebe0aae2a962c3371c88a547f6dc9fe475c61bd92b5e316
Literature & Architecture
Foundational Research & Citations
Primary scientific publications & architecture preprints
Gemma 4 Technical Report
Gemma Team, Google DeepMind
Presents the Gemma 4 open-weight model family, introducing multimodal dense and MoE architectures, native thinking reasoning traces, and unified encoder-free audio and image processing.
Gemma: Open Models Based on Gemini Research and Technology
Gemma Team, Google DeepMind
Foundational architecture, safety evaluations, and training methodology derived from Google Gemini research and technology.
Which Quantization Should I Use? A Unified Evaluation of llama.cpp Quantization
Empirical GGUF Research Study
Comprehensive empirical analysis of llama.cpp GGUF quantization schemes (Q8_0, Q4_K_M, Q2_K) across reasoning benchmarks, memory footprints, and CPU/GPU throughput.
Limitations
- RESEARCH_USE_ONLY — This model inherits a community license for research and free use; no commercial-serving claims are made.
- VRAM_ESTIMATE_NOTICE — Artifact size is disk usage, not a measured VRAM requirement. Deployment requires peak-memory profiling.
- COMPRESSED_DERIVATIVE — Pruned and re-quantized derivative of google/gemma-4-31b. Credit belongs to original authors.
Public release identity
- Release ID
- gemma-4-31b-gguf-v1
- Release version
- 1
- Published
- 2026-10-11T03:03:35.000Z
- Model ID
- Omega-Factory/gemma-4-31b-GGUF
- Base model ID
- google/gemma-4-31b
- Catalog family
- Not supplied
- Quantization rung
- Not supplied
- Supersedes
- Not supplied
- HF repository
- Omega-Factory/gemma-4-31b-GGUF
- HF repository URL
- https://huggingface.co/Omega-Factory/gemma-4-31b-GGUF
- HF revision
- 94881b2d021bf8e9a5518b4afc273bf70256f11c
- HF revision URL
- https://huggingface.co/Omega-Factory/gemma-4-31b-GGUF/tree/94881b2d021bf8e9a5518b4afc273bf70256f11c
- Release manifest SHA-256
- 4301ea4e14f8a032ba61b650742ea3b2f2e5192131336c5332c95fa4ce26b15e
- Catalog SHA-256
- e87595a7c9a5d0a6ae3662985e99356632737f2a673b773c938e4f93b214a99a
- Catalog generated
- 2026-10-11T12:45:00Z
- Commercial use
- DISALLOWED
- License URL
- https://ai.google.dev/gemma/terms