Marketplace · Decision (classify · yes/no · score)
nimble:9b
Call nimble:9b through one OpenAI-compatible API. It's served by 1 host on the Inpherio grid, each at a price it sets — your request goes to the cheapest healthy one, and fails over to the next if it's down.
At a glance.
from openai import OpenAI
client = OpenAI(base_url="https://inpherio.co.uk/v1",
api_key=INPHERIO_API_KEY)
# /v1/systemone (decision models are not in the OpenAI SDK)
r = requests.post("https://inpherio.co.uk/v1/systemone", headers={"Authorization": "Bearer KEY"},
json={"model": "nimble:9b", "state": "My order arrived broken.",
"questions": {"refund": {"instructions": "Is this a refund request?", "type": "choice",
"criteria": {"yes": "money back", "no": "anything else"}}}})nimble:9b in detail.
A 9B decision model from Bespoke Labs for fast, typed classification.
- Published by
- bespokelabs
- Parameters
- 9B
- Download size
- 9.5 GB (as packaged for Ollama)
- Quantisation
- Q8_0
- Format
- gguf
- Architecture
- qwen35
- Licence
- Apache 2.0
- Based on
- Qwen/Qwen3.5-9B
- Task
- text classification
- Library
- peft
- Languages
- en
- First published
- 18 Sep 2026
- Last updated
- 24 Sep 2026
- Downloads (30 days)
- 1,328 on Hugging Face
- Likes
- 251
peft lora qwen3.5 text-classification structured-prediction evidence-grounding en
The model card
Published by the model's authors on Hugging Face — shown as they wrote it, unedited by Inpherio.
Bespoke-Nimble-9B
Update (September 24, 2026): Updated to our latest 9B checkpoint, with an 8,192-token context limit and up to 255 choices per field. This release uses T=1.0; the earlier checkpoint remains available at revision="original-2676", and v2 is unchanged.
A Qwen3.5-9B LoRA adapter for context-grounded choices, booleans, and rubric score levels. It scores the allowed answer tokens directly; the evaluated workflow does not generate reasoning or free-form answers.
This repository contains the adapter (about 165 MiB), tokenizer, exact prompt builder, reference inference code, and release metadata. It requires the Qwen3.5-9B base checkpoint. The full base weights are not duplicated here.
License: Apache 2.0.
More info: https://github.com/bespokelabsai/nimble
Loading and scoring
Use a CUDA GPU with BF16 support. The original run used PyTorch 2.8.0 with CUDA 12.8. Install an appropriate CUDA build of PyTorch, then the remaining pinned requirements. The helper processes fields individually and preserves the training prompt and probability calculation.
hf download bespokelabs/Bespoke-Nimble-9B --local-dir nimble-model
pip install -r nimble-model/requirements.txt
Authenticate with a Hugging Face token that can read this repository if it is private. Review the included helper before importing it; no trust_remote_code=True is required.
import sys
sys.path.insert(0, "nimble-model")
from inference import NimbleModel
model = NimbleModel("nimble-model")
result = model.score(
context="The store accepts returns within 30 days. This item was bought 12 days ago.",
schema={
"eligible": {
"type": "boolean",
"description": "Is this item within the store return window?"
}
},
)
print(result["output"])
print(result["fields"]["eligible"]["probabilities"])
For choices, use type="enum", a choices list of strings, a description, and optional choice_descriptions. For rubric scores, use integer-valued enum strings such as ["0", "1", "2"], describe each level, and pass that field name in score_fields=["quality"]. This returns the selected integer and its probability-weighted expected score. Each field supports at most 255 choices. Prompts exceeding 8,192 tokens are rejected rather than truncated.
The included scorer applies softmax(candidate_logits / temperature) once, with a default of T=1.0. This checkpoint has not had a separate temperature fit; do not reuse the older model’s 2.179 calibration. For Mac/MLX and Linux loaders, use the current GitHub quickstart.
Sources: bespokelabs/Bespoke-Nimble-9B on Hugging Face · nimble:9b on Ollama. Collected from public sources on 8 Nov 2026; the host serving it may run a different build or quantisation — the marketplace shows each host's own measurements.
Who's serving it.
Every host benchmarked this model on its own hardware before serving it. Prices are set by each host; health is scored by the platform from what it observes.
| Host | Tier | Price | Measured speed | Health | State | Rent |
|---|---|---|---|---|---|---|
| 7B-A1B (5G) | L1 | £0.03 | 38344 ms | 82 | loads on request | £0.04/h |
Live data, refreshed on every visit. How billing works.
Room on the grid for your machine.
Right now one machine serves nimble:9b. If it drops, so does every request for it — a second host is what turns it into failover, and the grid needs that more than another copy. The router weighs price, measured speed, health and whether the model is already loaded — so a host with a sharper price or quicker hardware wins the next call.
Set your own price — you keep 85% of what it earns, and new hosts start with 12 months fee-free (up to £250 of earnings). A payout fee of £2 + 1% per payout to a UK bank account, £2 + 2% to any other country applies when you're paid. Install the connector, load nimble:9b, and the platform benchmarks it on your hardware before your first request.