AI Agent StoreDeep Principle · Gated checkpoint listed
Hydrogen 1.0
Explore scientific questions and draft research workflows for expert review. Access to the model files requires accepting the publisher’s conditions.
First reviewed 2026-11-03 · Checked 2026-11-03. The launch date has not been confirmed.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Try it on your hardware and tasks before relying on it.
What is announced so far
You download the model files and supply the hardware and software to run them. Capacity is the amount of information a request can contain; settings and available memory can limit what fits.
Exact download ID: deep-principle/Hydrogen-1.0. Post-trained from GLM-5.3 for chemistry, materials science and life-science questions. Published demonstrations use external tools and research environments. Neither the parent’s scores nor the publisher’s scientific results are independent measurements for this profile.
Hugging Face requires login and agreement to share account information before file access. The public card labels BF16 weights MIT and independently authored code/docs Apache 2.0; review the actual notices when obtaining access. Its serving example uses eight GPUs and custom code, and requires generation_config.json for correct answer stopping. Separate NVFP4/MXFP4 versions target different hardware.
The public card advertises 1M context and 128K output, subject to runtime resources and configuration. Exact file-level limits could not be verified without accepting the repository’s access conditions; the 4,096-token example is not the maximum.
No verified token tariff or measured local operating cost. The publisher also advertises NeuroSpark API and Mira access; those routes are not established by the Hugging Face inference-provider list.
Provider release and specifications ↗
Check availability and plans ↗
Test it before switching
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. Include the time and hardware needed to run it.
Compare caching and long-context costs ↗