Run ggml-org/models-moved locally
ggml-org/models-moved is a mid-size language model with 7.24 billion parameters, built on the llama architecture. It has been downloaded 598,823 times.
To run ggml-org/models-moved locally at a 4,096-token context, its quantized versions need between 1.1 GB (GGUF, lowest quality) and 8.25 GB (F16, highest quality) of memory, weights plus KV cache and a system margin included.
For most users the best balance is IQ3_S, needing about 3.79 GB. That means ggml-org/models-moved fits entirely in the VRAM of a 6 GB GPU or larger, running fully on the GPU.
All quantizations
| Quant. | Bits | Quality | Weights | KV | Total | Speed~ | Verdict |
|---|---|---|---|---|---|---|---|
| GGUF | 0.33 | Very low | 0.28 GB | 0.03 GB | 1.1 GB | 1438.2 t/s | Fits in VRAM |
| Q4_0 | 1.81 | Very low | 1.53 GB | 0.03 GB | 2.35 GB | 261.8 t/s | Fits in VRAM |
| Q8_0 | 3.36 | Fair | 2.83 GB | 0.03 GB | 3.66 GB | 141.4 t/s | Fits in VRAM |
| IQ3_S | 3.52 | Fair | 2.96 GB | 0.03 GB | 3.79 GB | 135.0 t/s | Fits in VRAM |
| F16 | 8.8 | Excellent | 7.42 GB | 0.03 GB | 8.25 GB | 6.7 t/s | Offload |
KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.
Frequently asked questions
How much VRAM do you need to run ggml-org/models-moved?
You need about 3.79 GB of VRAM to run ggml-org/models-moved entirely on the GPU using the IQ3_S quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run ggml-org/models-moved on an 8 GB GPU?
Yes. With 8 GB of VRAM you can run ggml-org/models-moved fully on the GPU using IQ3_S (about 3.79 GB).
Can I run ggml-org/models-moved on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run ggml-org/models-moved fully on the GPU using F16 (about 8.25 GB).
Can I run ggml-org/models-moved on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run ggml-org/models-moved fully on the GPU using F16 (about 8.25 GB).
What is the best quantization for ggml-org/models-moved?
If memory allows, higher bits-per-weight means better quality. A common sweet spot is a Q4_K_M or Q5_K_M quantization, which keeps most of the quality while roughly halving the memory versus 8-bit. Pick the highest quantization that still fits in your VRAM.