Caliber Flex 1B by OnePass
Decision-only model: no chat, no completions. You send a text plus questions, it returns picks with calibrated probabilities.
Exactly 1,000,000,000 parameters: Qwen3.5-0.8B-Base (Apache-2.0) depth-upscaled with 7 grafted top layers, trained LoRA r17, two-layer decision head (183→4392) with fitted temperature T=1.548. Counted, not rounded. Forecast reading taught by a TimesFM 3.0 teacher (non-commercial). English + Russian.
Files
| File | What |
|---|---|
caliber-flex-1b-Q8_0.gguf | Backbone + LoRA, Q8_0 quant. Generative/embedding use (chat, RAG). |
hybrid_adapter.pt | Decision head + LoRA + temperature. Needed for decisions. |
merged/ | Full bf16 transformers model (backbone + LoRA merged). |
infer.py | Standalone decisions, works offline. |
Modelfile | Ollama import for the GGUF. |
The GGUF alone does not answer decisions — the scoring head lives in
hybrid_adapter.pt and runs through infer.py (transformers).
The GGUF is the same backbone for open use: chat, embeddings, evals.
Decisions (transformers)
pip install torch transformers peft
python infer.py --state "Ticket #4411: two charges, refund one." \
--type choice --options billing shipping access
python infer.py --state "Тикет #1234: дважды списали деньги." \
--type choice --options billing shipping access
{
"model": "caliber-flex-1b",
"choice": "billing",
"probabilities": {"billing": 0.91, "shipping": 0.04, "access": 0.05},
"confidence": 0.87,
"route_to_human": false
}
Three question types: choice (one of N), noul (yes/no probability),
score (ordered levels, answered with the mean). No evidence in the
state → flat probabilities and "route_to_human": true.
| Metric | Value |
|---|---|
| Held-out accuracy | 0.800 |
| Brier | 0.074 |
| ECE | 0.037 |
| Temperature | 1.548 |
NoulXP bench (, home i5-4250U CPU, Q8_0, concurrency 1):