timestral cpp needs about 0.7 GB of memory at Q4_K (a 0.1 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 14 quants, from 0.0 GB (Q8_0) to 0.8 GB (F32). Longer context needs more: at 32K tokens the KV cache alone adds about 0.2 GB.
Timestral cpp 123M GGUF model for time‑series forecasting
Timestral cpp is a 123M parameter time‑series foundation model converted to GGUF for native CPU inference. It supports up to 32,768 time‑step context and 1,024‑step forecast horizon. The model outputs 21 quantiles for univariate or multivariate targets.
Good at
- Zero‑shot forecasting
- Large context length
- Quantile prediction
Limits
- No vision support
- License is Modified MIT
- No Python or libtorch required
Best for
- Forecasting with limited GPU memory
- Batch forecasting on CPU
- Time‑series anomaly detection
Who it’s for. Developers and data scientists needing a lightweight, zero‑shot forecasting model that runs natively on CPU. The model is suitable for production environments where Python dependencies are undesirable.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q8_0 | 0.0 GB | 0.6 GB | Fits | Get file → | |
| F16 | 0.0 GB | 0.6 GB | Fits | Get file → | |
| F32 | 0.0 GB | 0.6 GB | Fits | Get file → | |
| Q8_0 | 0.0 GB | 0.6 GB | Fits | Get file → | |
| F16 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| F32 | 0.1 GB | 0.7 GB | Fits | Get file → | |
| Q4_K | 0.1 GB | 0.7 GB | Fits | Get file → | |
| Q5_K | 0.1 GB | 0.7 GB | Fits | Get file → | |
| Q6_K | 0.2 GB | 0.7 GB | Fits | Get file → | |
| Q8_0 | 0.2 GB | 0.8 GB | Fits | Get file → | |
| F16 | 0.2 GB | 0.8 GB | Fits | Get file → | |
| F16 | 0.4 GB | 1.0 GB | Fits | Get file → | |
| F32 | 0.5 GB | 1.0 GB | Fits | Get file → | |
| F32 | 0.8 GB | 1.4 GB | Fits | Get file → |
For 8‑GB VRAM, use Q4_K (file size 0.12 GB, estimated 0.7 GB memory at 8k context). For 4‑GB VRAM, use Q5_K (0.15 GB, 0.7 GB memory). For 2‑GB VRAM, use Q6_K (0.17 GB, 0.7 GB memory). Estimates are approximate.
What GPU do you need for timestral cpp?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is F32 (0.8 GB file, about 1.4 GB in use).
Which timestral cpp quant should you choose?
Q4_K (0.1 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.
If memory is tight, the smallest file is Q8_0 at 0.0 GB, with the quality trade-offs that come with fewer bits. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about timestral cpp
How much VRAM does timestral cpp need?
About 0.7 GB at Q4_K with an 8K context window. That is the 0.1 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run timestral cpp on an 8 GB GPU?
Yes, with F32: the 0.8 GB file needs about 1.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of timestral cpp GGUF?
It depends on the quant: Q8_0 is 0.0 GB, F16 is 0.0 GB, F32 is 0.0 GB, Q8_0 is 0.0 GB, F16 is 0.1 GB, F32 is 0.1 GB and 8 more.
What context length does timestral cpp support?
The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.1 GB at 8K and 0.2 GB at 32K.
Where can I download timestral cpp GGUF?
The files are on Hugging Face in the mudler/timestral-cpp-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does timestral cpp use?
Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Timestral cpp on 8 GB VRAM?
Yes, using Q4_K or Q5_K quantizations keeps memory below 1 GB at 8k context.
Does this model support GPU inference?
Inference is native CPU only; GPU support is not available in the GGUF artifacts.
What license applies to this model?
The model is released under a Modified MIT license; see the license file for details.