Downloads · 30 days
128
14% of all-time downloads
gemma-4-E4B-it-oQ6-mtp
mlx-works/gemma-4-E4B-it-oQ6-mtp
gemma-4-E4B-it-oQ6-mtp is a machine learning model from mlx-works. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for mlx.
This model was quantized using oQ (oMLX v0.5.5) mixed-precision quantization.
All-time downloads
901
Public
Parameters
8.1B
7.4 GB on disk
Likes
1
Public
Repo makeup
Click a slice to open those files.
.safetensors7.4 GB · 100%
Parameter types
How the weights are stored.
U327.5B · 93%
Model card
From the Hugging Face model README
gemma-4-E4B-it-oQ6-mtp
This model was quantized using oQ (oMLX v0.5.5) mixed-precision quantization.
Base model: google/gemma-4-E4B-it
MTP head: google/gemma-4-E4B-it-assistant (merged into single checkpoint)
Quantization details
- Model type: gemma4
- Bits: 6
- Group size: 64
- Format: MLX safetensors
- MTP: Merged from assistant checkpoint (Lightning MTP)
Environment
- Hardware: M5 MacBook Air 32GB
- Inference Framework: oMLX v0.5.5
- Max Concurrent Requests: 4
- Settings:
- Thinking: Disabled
- TurboQuant KV Cache: Enabled (4-bit)
- Lightning MTP: Enabled (key speed improvement)
Performance Benchmarks
Note: Results are for reference only and may vary depending on hardware, software configuration, and workload.
Single Request Results
| Test | TTFT(ms) | TPOT(ms) | pp TPS | tg TPS | E2E(s) | Throughput | Peak Mem |
|---|---|---|---|---|---|---|---|
| pp1024/tg128 | 592.9 | 35.43 | 1727.2 tok/s | 28.5 tok/s | 5.113 | 225.3 tok/s | 6.37 GB |
| pp4096/tg128 | 2699.2 | 37.35 | 1517.5 tok/s | 27.0 tok/s | 7.465 | 565.9 tok/s | 6.45 GB |
Continuous Batching (pp1024 / tg128)
| Batch | tg TPS | Speedup | pp TPS | pp TPS/req | TTFT(ms) | E2E(s) |
|---|---|---|---|---|---|---|
| 1x | 28.5 tok/s | 1.00x | 1727.2 tok/s | 1727.2 tok/s | 592.9 | 5.113 |
| 2x | 52.1 tok/s | 1.83x | 1318.1 tok/s | 659.0 tok/s | 1553.7 | 6.464 |
| 4x | 97.1 tok/s | 3.41x | 1634.4 tok/s | 408.6 tok/s | 2328.0 | 7.779 |
Intelligence Benchmark
Note: Each benchmark round tests only 30 questions. Results are for reference only.
| Benchmark | Accuracy | Correct | Total | Time(s) | Think |
|---|---|---|---|---|---|
| MMLU | 70.0% | 21 | 30 | 24.3 | No |
| TRUTHFULQA | 60.0% | 18 | 30 | 9.3 | No |
| GSM8K | 86.7% | 26 | 30 | 130.8 | No |
| MATHQA | 23.3% | 7 | 30 | 69.2 | No |
| HUMANEVAL | 93.3% | 28 | 30 | 202.6 | No |
- Library
- mlx
- Type
- gemma4
- Created
- Aug 5, 2026
- Updated
- Aug 16, 2026