API-first decision model that turns text into calibrated probabilities for every allowed answer — yes/no, best choice, or grades — with no sampling.
Homepage hero — decisions, with the odds attached
1 / 3Overview
What is Nimble?
SOTA Decision Model
Nimble by Bespoke Labs is a state-of-the-art decision model: a hosted API that turns text into decisions. Send the state and the answers you allow in one request, and it returns a calibrated probability for every option — no sampling, no regex-parsing of LLM outputs. Three decision types are supported: noul for yes/no probability, choice for picking the best option with per-option probabilities, and score for grading with per-level probabilities. Nimble-v4 and Nimble-v4-max push the frontier on the Decision Index benchmark, scoring +10.5 and +5.74 over Jev 1.13, with internal evals showing 43% fewer mistakes than Jev and +50pp on model-routing decisions. Pricing is simple: input tokens only ($0.04 per million for nimble-v4, $0.09 per million for nimble-v4-max), output tokens free, paid from prepaid credit — and new accounts get a $1 free credit on email verification (about 3,000 decisions). Built for developers building AI agents and apps that need reliable text classification, routing and moderation decisions.
Platforms and languages
Nimble Availability
Platforms
Languages
Capabilities
Nimble Key Features
Calibrated probabilities per answer
One API request returns a calibrated probability for every allowed answer, eliminating sampling and regex-parsing of LLM outputs.
Three decision types
Supports noul (yes/no probability), choice (best option with per-option probabilities), and score (grade level with per-level probabilities).
SOTA decision accuracy
nimble-v4-max scores +10.5 over Jev 1.13 on the Decision Index benchmark, with 43% fewer mistakes than Jev in internal evals.
Input-tokens-only pricing
The hosted API charges for input tokens only ($0.04/M for nimble-v4, $0.09/M for nimble-v4-max) — output tokens are free.
Free credits to start
The web console offers free credits for testing, with a $1 credit grant on email verification (roughly 3,000 decisions).
Best for
Who uses Nimble?
Developers building AI agents and apps
make reliable text classification, routing and moderation decisions with calibrated probabilities
Plans and access
Nimble Pricing
Freemium
$0.04-$0.09 per million input tokens (output tokens free); $1 free credit on email verification.
Common questions
Nimble FAQs
What is Nimble?
Nimble is a hosted decision model by Bespoke Labs: send text and a set of allowed answers, and it returns a calibrated probability for every option — ideal for classification, routing and moderation.
How much does Nimble cost?
You pay for input tokens only: $0.04 per million for nimble-v4, $0.09 per million for nimble-v4-max. Output tokens are free. New accounts get a $1 free credit after email verification.
What decision types does it support?
Three: noul (yes/no probability), choice (best option with per-option probabilities), and score (grade level with per-level probabilities).
How accurate is it?
Nimble-v4-max scores +10.5 over Jev 1.13 on the Decision Index benchmark, and Bespoke Labs' internal evals report 43% fewer mistakes than Jev.
Reviews
See what the community thinks and share your experience.
—
Based on 0 ratings
Rating distribution
Community reviews
No reviews
No written reviews yet.
