timesfm vs clip
Two of the top projects in Open models, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
timesfm
Pretrained time-series foundation model for zero-shot forecasting from Google Research
clip
OpenAI's reference PyTorch code and weights for contrastive image-text embeddings
Verdict
timesfm ranks higher, 74 to 26. Of the 4 score parts we measure for both, timesfm leads on 4, clip on none.
- Adoptiontimesfm leads by 51 points, 78 to 27: a wide gap.
- Freshnesstimesfm leads by 41 points, 100 to 59: a wide gap.
- Maintenancetimesfm leads by 58 points, 58 to 0: a wide gap.
- Agent-readytimesfm leads by 30 points, 30 to 0: a wide gap.
Pick timesfm if you are
Engineers and researchers forecasting time series without per-task training
- Zero-shot forecasting of univariate and multivariate series with past-only and future covariates
- Returns point forecast plus 9 quantiles (0.1 to 0.9) per step
Pick clip if you are
ML engineers needing image-text embeddings or zero-shot classification
- Small API: clip.load, clip.tokenize, encode_image, encode_text
- Zero-shot classification from text labels without task-specific training
| What we compare | timesfm | clip |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 78, popular | 27, niche |
| Freshness | 100, active | 59, recent |
| Maintenance | 58, fair | 0, weak |
| Easy to run | Not scored for this kind | Not scored for this kind |
| Agent-ready | 30, minimal | 0, none |
| Facts from GitHub and the README | ||
| Stars | 34.5k | 34.5k |
| License | Apache-2.0 (permissive) | MIT (permissive) |
| Last commit | Sep 2026 | Mar 2026 |
| Last release | Aug 2026 | None published |
| Language | Python | Jupyter Notebook |
| Docker | No | No |
| GPU | Optional | Optional |
| arm64 or Apple Silicon | Mentioned | Not stated |
timesfm
TimesFM is a pretrained decoder-only model for time-series forecasting, installed as a Python package with PyTorch or MLX (Apple silicon) backends. Version 3.0 forecasts univariate and multivariate series with past-only and past-and-future covariates, returning point forecasts plus nine quantiles (0.1 to 0.9). The code is Apache-2.0, but 3.0 weights carry a non-commercial license.
Who it is for: Engineers and researchers forecasting time series without per-task training
Strengths
- Zero-shot forecasting of univariate and multivariate series with past-only and future covariates
- Returns point forecast plus 9 quantiles (0.1 to 0.9) per step
- MLX backend runs 3.0 on Apple silicon without PyTorch, numerically matched to PyTorch
- Includes LoRA fine-tuning example (HF Transformers + PEFT), unit tests and an agent SKILL.md
Weaknesses
- TimesFM 3.0 weights are non-commercial and non-production; commercial use only via Google Cloud services
- Not an officially supported Google product
- Forecasting only; contexts beyond 15,360 points are truncated to the most recent
- README code imports differ between 3.0 examples and the PyPI 2.0.2 note, which is unclear
- GPU optional
- Needs PyTorch or MLX, Hugging Face checkpoints
- Models: TimesFM 3.0 (PyTorch checkpoint), TimesFM 2.5 (200M params, 16k context), TimesFM 1.0 and 2.0 (v1 archive)
clip
CLIP is a Python package that loads OpenAI's pretrained image-text models and exposes encode_image, encode_text and a similarity forward pass. It supports zero-shot image classification from natural-language labels and feature extraction for linear-probe training. The README targets PyTorch 1.7.1 or later on CUDA or CPU, with weights downloaded on demand.
Who it is for: ML engineers needing image-text embeddings or zero-shot classification
Strengths
- Small API: clip.load, clip.tokenize, encode_image, encode_text
- Zero-shot classification from text labels without task-specific training
- MIT license; runs on CPU or CUDA GPU
- README has runnable zero-shot and linear-probe examples on CIFAR-100
Weaknesses
- Inference and evaluation only; no training code in this repo
- Installed from git via pip; no tagged releases
- Text context length defaults to 77 tokens
- Larger independently trained models are in OpenCLIP, not here
- GPU optional
- Needs PyTorch 1.7.1+, torchvision, ftfy, regex, tqdm
- Models: CLIP ViT-B/32, other names via clip.available_models()
More in Open models
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.