- Family: ARC-Neuron
- Tier: Tiny
- Parameters: ~0.05M
- Format: GGUF (minimal subset emitted by ARC Tiny export path)
- Precision: F32
- Architecture tag:
arc_tiny
ARC-Neuron Tiny is the bootstrap model for the ARC-Neuron line. It exists to prove the complete local path:
- train
- export GGUF
- validate GGUF
- load GGUF
- generate locally
- regression baseline
- container-format validation
- runner smoke tests
- documentation of the ARC local model path
- serious production use
- broad reasoning claims
- large-context coding work
- benchmarking against real general-purpose LLMs
- Primary artifact:
artifacts/gguf/ARC-Neuron-Tiny-0.05M-v0.1-F32.gguf - Legacy alias retained:
artifacts/gguf/ARC-Tiny-Demo-0.05M-v1.0-F32.gguf
python3 scripts/lab/validate_arc_tiny_gguf.py
python3 scripts/lab/run_arc_tiny_gguf.py --tokens 16ARC-Neuron Tiny is honest infrastructure proof, not a flagship intelligence claim.