Model Detail Human reviewed

Bespoke Nimble 9B

Bespoke Nimble 9B is an open typed-decision model fine-tuned from Qwen3.5-9B. It scores user-supplied enum and boolean answers directly, supports up to 255 choices per field and an 8K request limit, and can run on Apple Silicon or NVIDIA GPUs through the project's local serving paths. Its published evaluation is narrow and partly synthetic, so probability thresholds should be calibrated on the target workflow.

Bespoke LabsBESPOKE NIMBLEapache-2.02026-09-17

Deployment and license note

Decision models return bounded choices, scores, or yes/no probabilities instead of long-form text. Public benchmark results are useful for screening candidates, but production thresholds should be calibrated on a labelled set from your own routing, guardrail, or action-selection workflow.

Parameters
9B
Dense model
Context window
8K
Maximum decision input
Architecture
qwen3.5 direct logit decision
dense
Quality score
88
Planner signal

Task Fit

Typed DecisionsSupported

Bounded routing, classification, scoring, guardrails, and action selection with structured probabilities.

Code AgentNot a fit

Not marked for code agent in the current library.

CodeNot a fit

Not marked for code in the current library.

ChatNot a fit

Not marked for chat in the current library.

RAGNot a fit

Not marked for rag in the current library.

VisionNot a fit

Not marked for vision in the current library.

Image GenerationNot a fit

Not marked for image generation in the current library.

Video GenerationNot a fit

Not marked for video generation in the current library.

VoiceNot a fit

Not marked for voice in the current library.

Source Confidence

Overallhigh · 100/100
ParametersReviewed / seeded
Task fitReviewed / seeded
MemorySeeded artifact
LicenseSource / seed
BenchmarksAvailable
Hardware fitCalculated

Variants and Quant Artifacts

Choose the artifact first; hardware fit follows from RAM, VRAM, format, and runtime.

1 artifacts
QuantFormatQualityMin RAMReco RAMRuntimeAction
BF16safetensorshigh24GB32GBtransformers, mlx, sglang Plan with this

Benchmarks

Nimble Author Heldout Accuracy90.1%

Source and Review

Hugging Facebespokelabs/Bespoke-Nimble-9B
OllamaNot mapped
VerificationHuman reviewed
Artifact sourceofficial-huggingface-weights
Default variantBespoke Nimble 9B Decision
Tool callingNot marked

Execution evidence

Run Bespoke Nimble 9B with a documented recipe

Recipes connect hardware, a model artifact, tools, settings, verification, and a reportable result.

Browse all recipes →

No verified recipe is linked to this record yet.

Compatibility estimates remain available in the planner. A recipe appears here only after its exact stack and verification protocol are documented.

Continue planning your local AI setup