mars

<div align="center">

Mars

A code & design specialist that punches far above its weight.

HuggingFace · The Imagination Company · Built by Yuvraj Varma — independent researcher

</div>

---

Mars is a deliberately narrow Mixture-of-Experts model trained on only three things — English, software, and design. No math, no general-web sludge. Every parameter is spent where it counts.

The bet is simple: a small, sharp specialist that answers many times cheaply and shows you only the verified-best attempt can rival models a hundred times its size — on the tasks Mars is built for. At ~1B active parameters, Mars can afford 50 tries for the price of one frontier response, and code and design both have mechanical verifiers to pick the winner.

<table width="100%"> <tr> <td width="50%" valign="top">

Mars 8B (Mars-Mini)

An 8.04B-parameter MoE (48 experts, top-2 → ~1.09B active per token) trained from scratch on a two-GPU workstation. The experts hold ~7.2B of capacity; only ~0.3B of it fires per token — capacity of a big model, cost of a small one.

Architecture

• MoE — 48 fine-grained experts, top-2 routing + load-balancing
• MLA — Multi-head Latent Attention (compressed KV cache)
• MTP — Multi-Token Prediction (2 tokens ahead)
• 24 layers · d 2048 · 16 heads · expert FFN 1024
• RoPE θ = 10M → long-context headroom
• 131,072 vocab · stochastic-rounded bf16

Trained with

• Aurora — leverage-aware spectral optimizer (beats Muon)
• TST — Token-Superposition Training (~2.5× data-efficiency)
• Sol Tokenizer — code-heavy BPE

</td> <td width="50%" valign="top"> </td> </tr> </table>

---

The Roadmap

Mars goes through a code-and-design-first post-training stack:

Pretrain (TST) → SFT (tests-first + generate→run→repair) → RLVR (executed tests · design audits) → Best-of-N inference scaffold

The reward that matters is mechanical: for code, unit tests pass in a sandbox (with partial credit for fraction-of-tests-passed and compile success); for design, automated audits on rendered output. A signal that doesn't care how big the model is.

---

<div align="center">

Beating a frontier model with 8B parameters isn't a weights problem — it's a product problem. Mars is a bet that MoE for capacity, MLA for memory, MTP for speed, TST for training efficiency, and a verify-and-repair scaffold for inference can make a specialist go far above its weight — in code and design.

Mars · The Imagination Company

</div>