ZeroS-Pico-v1.0

ZeroS-Pico-v1.0 is a 67K-parameter small language model (SLM) featuring a custom architecture inspired by Needle2. Trained entirely on CPU over 600M tokens, ZeroS-Pico-v1.0 utilizes a byte-level tokenizer with 4 special tokens (<bos>, <eos>, <pad>, <unk>), resulting in a compact vocabulary size of 260.

Architecture

ZeroS-Pico-v1.0 employs a compact, parameter-efficient architecture incorporating Engram memory, Hadamard FFNs (parameter-free) with SwiGLU intervals, and an 8-stream topology powered by mHC.

• Vocab Size: 260
• Max Position Embeddings: 96
• Hidden Size: 32
• Intermediate Size (for SwiGLU): 64
• Total Number of Layers: 9
• Hadamard Layers: 7
• SwiGLU Layers: 2
• Number of Heads: 4
• Number of KV Heads: 2
• Dimensions Per Head: 8
• Use Per-Head Gating: false
• Use XSA: false
• Number of mHC Streams: 8
• Use Engram: true
• Number of Engram Entries: 196
• Engram Orders: (4, 8)

Training Dataset

ZeroS-Pico-v1.0 was trained on 600 million tokens of a diverse dataset mixture comprising general web text, educational content, synthetic data, normalized code, and mathematics.

Benchmark Results

We benchmaked ZeroS-Pico-v1.0 on five tasks: Arc_Easy, Arc_Challenge, HellaSwag, PiQA, and ArithMark-3.0.

Despite its compact size, ZeroS-Pico-v1.0 exhibits surprisingly competitive performance on knowledge-intensive and mathematical benchmarks within its parameter class.

License

Apache 2.0.

Citation

@misc{zeros-pico-v1.0,
  title        = {ZeroS-Pico-v1.0},
  organization = {FromZero},
  authors      = {Paul Courneya},
  year         = {2026},
  url          = {https://huggingface.co/fromziro/ZeroS-Pico-v1.0}
}