shahjenil

Docket-3B

README

License: apache-2.0

Measured quality

Evaluated in bf16 with development-fitted calibration. Enterprise accuracy weights ten workflows equally; reasoning weights five tasks equally. The Jev reference scores 86.4% on enterprise decisions and 198/231 on JevBench.

Table with columns: Evaluation, Docket-3B
Evaluation	Docket-3B
Enterprise decisions /3,063	90.1%
Held-out reasoning /743	78.2%
Logical verification /160	100.0%
JevBench /231	177
Reasoning ECE	0.0536
Reasoning wrong at ≥0.9 confidence	0.54%

Wrong at ≥0.9 is the fraction of all labelled answers that are both wrong and at least 90% confident. No JevBench item was used in training or development-time adapter selection. Docket-3B accepts an 8,192-token serving context; vLLM reserves one answer token. This context limit does not establish accuracy at every input length. See the full benchmarks for test conditions, sampled public-dataset results, missing measurements, and limitations.

Serving

Serve it with docket-serve on Linux with an NVIDIA GPU (12 GB or more). vLLM compiles kernels at startup, so it needs a C compiler and the Python headers as well as the venv module; stock Ubuntu and cloud GPU images lack them:

bash

sudo apt-get update && sudo apt-get install -y python3-venv python3-dev gcc   # Ubuntu/Debiangit clone https://github.com/shah-jenil/docket-serve.git docket && cd docket./install.sh vllm                 # 3-8 minutes./serve.sh 3b                     # http://127.0.0.1:8080; first start 5-7 minutes (download, CUDA graphs)python3 scripts/smoke_test.py     # in a second terminal: PASS

Later starts take about 4 minutes on an RTX 3060 (CUDA graph capture), less on larger GPUs. Timings, stopping and troubleshooting: GPU guide.

Use the merged weights and their bundled docket_calibration.json. Calibration: {"choice": 1.15, "noul": 1.3, "score": 1.1}. The default configuration is bf16, matching the quality evaluation. FP8 inference is supported, but FP8 accuracy, latency and serving economics have not been remeasured for these selected weights. Published H100/L40S serving runs are reference studies. Request shapes, endpoints and limits: API reference.

Which release to use

Table with columns: Repo, Format, Use it for
Repo	Format	Use it for
Docket-3B	BF16 safetensors	GPU serving with vLLM; reference quality
Docket-3B-GGUF	GGUF Q8_0	Mac (Metal), CPU or a small GPU with llama.cpp

Licence and attribution

Apache-2.0. A derivative of IBM Granite 4.2 (Apache-2.0). Public data credits: WANLI and banking77 (CC BY 4.0), GSM8K (MIT), AQuA-RAT (Apache-2.0). No outputs of TypeSafe or OpenAI models were used as training data. Docket is not affiliated with or endorsed by IBM or TypeSafe AI. See NOTICE.

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Model Details

Model Provider

shahjenil

Model Tree

Base

ibm-granite/granite-4.2-3b

Fine-tuned

this model

Input Modalities

Text

Output Modalities