Fiber Pro — North ML Flare-1 (~1.6B)
The flagship Flare-1 checkpoint: Qwen3.5-derived gated DeltaNet + GQA
hybrid backbone (hidden 2048, larger MLP) with AgentRegisters, tuned
with a fact-check / self-correction RL pass on top of Fiber Lite. This is
the full-precision training checkpoint (HF transformers format). The
production 8-bit Kestrel build with the learned router, Fiber memory
trunk, and search pipeline is at
North-ML1/Fiber-pro.
Benchmarks (measured, same harness — Modal L4, greedy, oct 2026)
| Model | ARC-C | MMLU | GSM8K | HumanEval |
|---|---|---|---|---|
| ink (725M) | 63.8% | 32.0% | 30.0% | 37.5% |
| fiber-lite (1.6B) | 72.9% | 38.0% | 16.0% | 37.5% |
| fiber-pro (1.6B) | 80.9% | 40.0% | 14.0% | 48.4% |
| Qwen3-0.6B (raw) | 49.2% | 29.0% | 4.0% | 4.7% |
| Qwen3-1.7B (raw) | 72.9% | 39.5% | 43.0% | 68.8% |
| SmolLM3-3B (raw) | 77.9% | 36.0% | 51.0% | 71.9% |
Subsets: ARC-C n=199, MMLU n=200, GSM8K n=100, HumanEval n=64. Honest read: best knowledge-MC scores in this lineup — beats SmolLM3-3B at nearly 2× its size on ARC-C and MMLU. Trails the newer models on math (GSM8K 14%) and code; those are the known weak spots.
Specs
| Parameters | ~1.63B (1.41B backbone + 158M Fiber trunk + registers + bridge) |
| Layers | 24 |
| Hidden | 2048 |
| Attention | GQA 8 heads / 2 KV, DeltaNet:GQA = 3:1 |
| MLP | 6144 |
| Context | 262,144 tokens |
| Vocab | 65,536 |
Interactive chat
pip install torch transformers safetensors huggingface_hub
python3 chat.py --model North-ML1/fiber-pro
chat.py (in this repo) auto-downloads the weights, attaches the
AgentRegisters, and starts a REPL with optional two-phase thinking.
Links
- Architecture report: Flare-1 paper
- Runtime artifacts: Ink-1 · Fiber · Fiber-pro
- Downloads last month
- 42