Skip to content
Other

Atlas-5K 25M

weather forecasting model

1 min readupdated

Atlas-5K 25M from ParadoxAI Lab is built for 24-hour weather forecasting. It predicts station-level temperature, dew point, wind speed and direction, and sea-level pressure. Built around the WEATHER-5K dataset, the model packs roughly 24 million parameters. This release scales up the previous 10M model and outperforms competing approaches across several benchmarks. It is geared toward researchers and developers working in meteorology and time-series forecasting (predicting sequences of data that change over time).

Key features

  • Trained on the WEATHER-5K dataset, covering 5,672 weather stations worldwide from 2014 to 2023.
  • Contains 24,103,796 parameters with a scaled-up hidden layer architecture.
  • Outperforms PhysicsFormer and several other forecasting systems.
  • Supports loading weights and metadata via the Hugging Face Hub library.

What you can do

  • Forecast temperature, pressure, and wind 24 hours ahead using weather station data.
  • Integrate the model into time-series pipelines using the Paradox World Model repository.

Example

import torch
from huggingface_hub import hf_hub_download
path = hf_hub_download("paradoxailab/atlas-5k-25m", "59_atlas-5k-25m_weather5k_adopted.pt")
ck = torch.load(path, map_location="cpu", weights_only=False)
print(ck["meta"])            
weights = ck["weights"]       
results = ck["results"]       
Read moreDetails
Benchmarks: validation E 0.27307, 2023 test MAE T 1.400 / Td 1.468 / WD 53.33 / WS 1.173 / SLP 1.473.
Training: Google Colab (A100 and L4).
License: Apache-2.0.
Discuss Telegram 1

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # paradoxailab/atlas-5k-25m
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download paradoxailab/atlas-5k-25m --local-dir ~/models/atlas-5k-25m
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to atlas-5k-25m
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of atlas-5k-25m (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

paradoxailab/atlas-5k-25m
Author
paradoxailab
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments