wowsuch ai
Live

Bespoke LabsModel

Bespoke-Nimble-9B-v2

Bespoke-Nimble-9B-v2 is a PEFT LoRA adapter for text-based, schema-conditioned decisions, requiring the separately downloaded Qwen/Qwen3.5-9B base model. It scores allowed answer-token logits to provide boolean judgments, categorical choices, ordinal scores, and probability distributions; it does not generate reasoning or free-form answers.

What it does

Use it to score structured decisions against supplied choices and obtain probability distributions for those choices.

Availability
Download the complete package from Hugging Face.
Pricing model
Not established
Deployment
Requires a separately downloaded Qwen3.5-9B base model; reference scorer uses CUDA.
Privacy
Not established
Licence
Package includes a license; the license name is not stated.
Capability
analysisextraction
Modality
text
Access
open weights
Deployment
on-premises
Pricing
unknown
Target user
developerresearcherenterprise

Evidence

6 claims
  • The adapter scores allowed answer-token logits for boolean judgments, categorical choices, and ordinal scores. [1]
    boolean judgments, categorical choices, and ordinal scores with probability
    Bespoke Labs. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The included inference code supports 1–255 enum choices per field. [1]
    The included `inference.py` now supports 1–255 enum choices per field.
    Bespoke Labs. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The adapter is approximately 173 MB and requires the separately downloaded Qwen/Qwen3.5-9B base model. [1]
    The approximately 173 MB adapter requires the separately downloaded base model:
    Bespoke Labs. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The reference scorer requires Python 3.11 and a CUDA GPU supporting BF16. [1]
    Use Python 3.11 with a CUDA GPU supporting BF16 and enough memory for the 9B base
    Bespoke Labs. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The 255-choice limit describes runtime support, not measured accuracy or calibration at every candidate count. [1]
    The 255-choice limit describes runtime support, not measured accuracy or
    Bespoke Labs. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The adapter does not generate reasoning or free-form answers. [1]
    It does not generate reasoning or free-form answers.
    Bespoke Labs. Quoted text is outside CC BY 4.0 and keeps its original rights.

Sources

1
1
Bespoke LabsOwner-controlled
https://huggingface.co/bespokelabs/Bespoke-Nimble-9B-v2 Retrieved 2026-10-28 00:02 UTC
Verified

Versions

3
VersionEffectiveChangeChanged fieldsState
32026-10-24 01:40 UTCevidence updatescoreCurrent
22026-10-16 21:35 UTCmaterial changeavailability, deploymentCurrent
12026-10-15 22:34 UTCpublicationNoneCurrent

In briefs

2

Open in the catalog

Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.