Model reference · open weights

voilum-1

Available as managed deployment LLMs elisepaul · community Text gen 1 variants 536 dl/mo

voilum-1 is an open-weight language model from elisepaul. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.

Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.

What it is

Released byelisepaul
TypeLanguage models
TaskText gen
Parameters (lead)464M
Runs withpytorch
Released2026-09-12
Popularity536 downloads / month
LicenceOpen weights

About

What voilum-1 is


Abstract

We introduce Voilum-1, a high-efficiency 463.6-million parameter (100.8M active parameters per token) sparse Mixture-of-Experts (MoE) Small Language Model (SLM) specialized for deterministic, mathematically rigorous Python code synthesis, numerical scientific computing, and deep learning systems implementation. Trained across 12.0 billion tokens of curated algorithmic and mathematical data, aligned via ChatML instruction tuning, and reinforced through a 6-tier curriculum of execution-grounded Group Relative Policy Optimization (GRPO), Voilum-1 solves complex machine learning tasks from conceptual specifications without code spoon-feeding. In empirical sliding-window evaluations, Voilum-1 achieves a 99.3% greedy pass rate on challenging applied machine learning and numerical systems benchmarks while operating within an ultra-compact memory footprint of 884 MB in BF16 SafeTensors.


1. Architectural Specifications

Voilum-1 employs a decoder-only sparse Mixture-of-Experts architecture incorporating Grouped-Query Attention (GQA), Rotary Position Embeddings (RoPE), RMSNorm, and SwiGLU gating:

Read the full model card
Architectural HyperparameterSpecificationDescription
Total Parameters463,553,024 (463.6M)Full weights memory footprint (884 MB in BF16)
Active Parameters / Token100,800,000 (~100.8M)Parameters engaged in each forward pass
Transformer Layers ( (N_{\text{layers}}) )16Pre-norm decoder blocks
Hidden Dimension ( (d_{\text{model}}) )512Model channel representation dimension
Feed-Forward Routing8 Experts (Top-1 / Top-2)Sparse token-level gating with softmax load balancing
Expert Intermediate Dim2,048SwiGLU projection dimension per expert
Attention MechanismGrouped-Query Attention (GQA)8 Query heads, 2 Key-Value heads (4:1 compression)
Head Dimension ( (d_k) )64Per-head projection dimension
Positional EncodingRotary Embeddings (RoPE)Base frequency ( \theta = 10,000.0 )
NormalizationRMSNormZero-mean centering omitted for scale invariance: ( \epsilon = 10^{-6} )
Activation FunctionSwiGLUGated Swish-linear non-linearity
Context Length ( (L_{\text{ctx}}) )2,048 tokensNative autoregressive training horizon
Vocabulary Size ( (V) )49,152Byte-Pair Encoding (StarCoder2 BPE)

Layer Dataflow & Computation

Each decoder layer executes the following standardized pre-norm transformation sequence:

  1. Pre-Attention Normalization: Input representation ( x ) is normalized via Root Mean Square Normalization:

$$ \text{RMSNorm}(x, \epsilon=10^{-6}) = \frac{x}{\sqrt{\frac{1}{d}\sum_{i=1}^{d} x_i^2 + \epsilon}} \odot \gamma $$

  1. Grouped-Query Attention (GQA): 8 query heads interact with 2 shared key-value heads (4:1 KV compression), rotated with RoPE position embeddings (base ( \theta=10,000.0 )). The resulting attention context is added to ( x ) via residual connection:

$$ \text{Attention}(Q, K, V) = \text{softmax}\left(\frac{Q K^T}{\sqrt{d_k}}\right) V $$

  1. Pre-FFN Normalization: The attention output is normalized through a second RMSNorm layer.

  2. Sparse Mixture-of-Experts Feed-Forward: A learnable routing gate computes softmax probabilities over 8 experts. The top-2 activated experts evaluate independent SwiGLU transformations with intermediate projection dimension 2,048:

$$ \text{SwiGLU}(x) = \left(\text{Swish}(x W_{\text{gate}}) \otimes (x W_{\text{up}})\right) W_{\text{down}} $$

The routed expert outputs are dynamically weighted by routing probabilities, summed, and added to the residual stream.


2. Training Methodology & Curriculum Progression

Voilum-1 was developed through a three-stage progressive optimization pipeline:

  • Stage 1: Base Pretraining (12.0 Billion Non-Repeating Tokens) Pretrained on a highly curated corpus spanning scientific Python (NumPy, SciPy, Pandas), PyTorch deep learning internals, algorithms, and numerical mathematics. Optimized with AdamW (peak learning rate ( 5 \times 10^{-4} ), linear warmup, cosine decay, bfloat16 mixed precision).

  • Stage 2: Direct Instruction SFT Alignment Aligned using ChatML formatting (/) under a strict Zero-Thinking Token policy. The model is trained to produce clean, direct, executable code without conversational boilerplate or unstructured internal monologue.

  • Stage 3: Pedagogical T-GRPO Reinforcement Learning (Tiers 1 to 6) Post-trained using Group Relative Policy Optimization (GRPO) with signed advantage baselines (( G = 4 ) rollouts per prompt) and KL penalty against the Stage 2 reference policy. Evaluated via an automated, isolated compiler and unit-test execution sandbox across 6 pedagogical mastery tiers.

2.1 The Zero-Code Authentic Problem Solving Standard (Tier 6)

Unlike conventional coding benchmarks that spoon-feed mathematical equations directly inside the prompt (e.g. rms = torch.sqrt(torch.mean(x**2) + eps)), Voilum-1 is reinforced against authentic engineering and applied system specifications:

  • The Problem Framing: Outlines the theoretical motivation, invariant properties, and mathematical constraints.
  • The Interface Specification: Defines expected function/class signatures, input/output tensor ranks, and data types.
  • Zero Implementation Dictation: The model autonomously formulates algorithmic transformations, dimensional broadcasting, and vectorized execution from first principles.

2.2 Curriculum Benchmarks Across All 6 Mastery Tiers

All tiers were evaluated under strict deterministic greedy decoding (( T = 0.0, \text{argmax} )) over rolling evaluation windows of ( W = 15 ) (300 training steps):

$$ \text{Rolling Greedy Pass Rate} = \frac{1}{W} \sum_{k=1}^{W} \text{GreedyEval}_k \ge 25.0% \implies \text{Tier Mastered} $$

| Curriculum

From the published model card. Full card on the HuggingFace links in the sidebar.

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys voilum-1 for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (voilum-1 below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/chat/completions \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"voilum-1","messages":[{"role":"user","content":"Hello"}]}'

Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms