Indian Flag
Government Of India
A-
A
A+
ORGANISATION

LUCAN 10B

LUCAN 10B Physics MoE is an 11B parameter multi-domain scientific foundation model built on a hybrid Mamba2 state-space and 12-layer Sparse Mixture-of-Experts (MoE) architecture with a PhysicsConstrainedOutputHead. It models continuous scientific latents across fluid transport, chemical reactors, phase-change thermal physics, and molecular systems.

About Model

LUCAN-10B-Physics-MoE is a multi-domain scientific foundation model developed by Shodh AI Private Limited. The model includes a native PhysicsConstrainedOutputHead (TCPI) enforcing non-negativity and integral mass/flux conservation directly inside the forward pass.

Multi-Domain Scaling Over LUCAN-3B: Achieves a 9.86% reduction in validation Relative L2 error (1.1933 -> 1.0756) and a 6.01% reduction in held-out test Relative L2 error (1.1682 -> 1.0980), improving across all 5 physics domain groups (Compressible Fluids -12.71%, Reactors & Erosion -6.18%, 3D Phase-Field Solidification -14.26%, Molecular Conformations -7.34%, and Aerodynamic Shock Benchmarks -10.01%) with zero dropped tokens.

Inference Latency & Speedup (1× NVIDIA H100 80GB): Delivers a 37.63 ms median single forward pass latency, 5.98 seconds end-to-end execution across a 1,000-scenario parameter sweep (5.98 ms/case), and a 39.83× end-to-end speedup (0.0801 s vs. 3.1914 s) over the numerical Stefan phase-change solver.

LUCAN 10B

Metadata Metadata

Other

Shodh Ai research Team

Token Classification Model

PyTorch

Restricted

SHODH AI PRIVATE LIMITED

Science, Technology and Research

05/11/26 06:54:45

Ellwil Sharma

4.02 GB

compact_codec.py ( 16.18 KB )


To preview this file, you need to be a registered user. Please complete the registration process to gain access and continue viewing the content.

Activity Overview Activity Overview

  • Downloads0
  • Downloads 0
  • File Size 4.02 GB
  • Views 15

Tags Tags

  • Sovereign AI

License Control License Control

Other

Version Control Version Control

FolderVersion 1(4.02 GB)
  • Ellwil Sharma· Today
    • chevron_rightFolder
      LUCAN_10B_COMPACT_AI_KOSH
      • undefined
        compact_codec.py
      • application/json
        compact.index.json
      • undefined
        components-0001.safetensors
      • undefined
        components-0002.safetensors
      • undefined
        components-0003.safetensors
      • undefined
        components-0004.safetensors
      • undefined
        components-0005.safetensors
      • undefined
        components-0006.safetensors
      • undefined
        components-0007.safetensors
      • undefined
        components-0008.safetensors
      • more_horiz 19 more

More Models from SHODH AI PRIVATE LIMITED More Models from SHODH AI PRIVATE LIMITED

LUCAN 10B
LUCAN 10B Physics MoE is an 11B parameter multi-domain scientific foundation model built on a hybrid Mamba2 state-space and 12-layer Sparse Mixture-of-Experts (MoE) architecture with a PhysicsConstrainedOutputHead. It models continuous scientific latents across fluid transport, chemical reactors, phase-change thermal physics, and molecular systems.
Sovereign AI
  • See Upvoters0
  • Downloads0
  • File Size4.02 GB
  • Views15
Updated Today

SHODH AI PRIVATE LIMITED

LUCAN 3B
LUCAN 3B is an approximately 3-billion-parameter scientific AI model from Shodh AI for processing numerical reactor data. This release includes the model weights, dedicated tokenizers and matching adapters for bioreactor and continuous-flow reactor applications, along with loading instructions and an introductory guide.
Sovereign AI
  • See Upvoters0
  • Downloads0
  • File Size4.44 GB
  • Views28
Updated 6 day(s) ago

SHODH AI PRIVATE LIMITED