Typed decision outputs
Send a state and the answers or levels you allow, then receive the selected answer together with a probability for each option.
Nimble
Nimble is a hosted API that turns text into typed decisions, returning an answer and probabilities for each allowed option. It is designed for workflows such as routing, tagging, grading, policy checks, and agent action gating.
Nimble decision models
Nimble is a hosted API for turning text into typed decisions. You provide a state—such as a customer message—and one or more questions with permitted answers. Nimble returns an answer and a probability for every option, allowing software to proceed with confident results and route uncertain cases for human review.
The product offers Nimble-v4 and Nimble-v4-max through the same request format. The site positions the 9B Nimble-v4 for high-volume routing, tagging, and triage, and the 27B Nimble-v4-max for policy checks, grading, and agent decision steps. Documented question types include truth-probability checks, named-choice classification, and ordered scoring.
Features
Noul questions
Ask whether a statement is true and receive its estimated probability on a 0-to-1 scale.
Choice questions
Name the possible answers for a classification task and receive the chosen option plus probabilities for the available choices.
Score questions
Provide ordered levels from lowest to highest and receive an expected level with a probability distribution across those levels.
Multi-question requests
Place multiple questions in one request. The site says Nimble reads the state once and returns an answer for every question.
Two model sizes
Use Nimble-v4 at 9B for high-volume routing, tagging, and triage, or Nimble-v4-max at 27B for policy checks, grading, and an agent's next step.
Use cases
Support ticket routing
Classify incoming tickets into departments such as billing, technical, account, or sales, then use the returned probabilities to send confident cases directly to a queue and review ambiguous ones.
High-volume tagging
Apply consistent labels to text at scale when the workflow needs a selected category and a confidence signal rather than an unconstrained text response.
Policy checks
Evaluate whether a request or response meets a stated policy, using the larger model for checks where the decision has greater operational importance.
Response grading
Score a reply against ordered levels such as a 1–5 scale and retain the distribution for review or downstream quality analysis.
Agent action gating
Use a decision result as a gate in an automated agent workflow, such as deciding whether the agent should take its next step or pause for a person.
Pros and Cons
Pros
- Returns probabilities alongside decisions, which supports confidence-based routing and human escalation.
- Supports multiple typed question formats and multiple questions in one request.
- Offers a smaller model for high-volume work and a larger model for more consequential checks, using the same request format.
- Provides a documented API request example and an email-verification offer of credits worth 3,000 decisions.
Cons
- The available evidence does not document integrations, SDKs, deployment options, authentication beyond the example bearer-key request, or operational limits.
- The pricing page supplied for review returned a 404, so pricing, quotas, and paid-plan details cannot be confirmed.
- Published benchmark figures are based on a public-suite run described as not yet being on the Decision Index board; the site says that suite represents 20% of the board's full score.
FAQ
What is Nimble?
Nimble is a hosted decision-model API. A request includes a state and one or more typed questions, and the response provides an answer plus probabilities for the available options.
How does a Nimble request work?
The documented quickstart sends a request to the Bespoke Labs API with a bearer API key, a model name, a state, and a questions object. The supplied example uses the nimble-latest model identifier.
What question types does Nimble support?
Nimble supports Noul questions for estimating whether a statement is true, choice questions for selecting among named options, and score questions for placing an item on ordered levels.
Which Nimble model should I use?
Nimble-v4 is positioned for high-volume routing, tagging, and triage, while Nimble-v4-max is positioned for policy checks, grading, and an agent's next step. Both use the same request format and return probabilities.
What pricing or starting credits are documented?
The site says users can start with credits worth 3,000 decisions after signing up and verifying an email. The supplied pricing page returned a 404, so no additional pricing, quota, or paid-plan details are documented here.
Quick Facts
- Category
- AI decision models / hosted API
- Models
- Nimble-v4 (9B) and Nimble-v4-max (27B)
- Question types
- Noul, choice, and score
- Output
- An answer and probability distribution for each question
- Access
- Hosted through the Bespoke Labs API; signup shown on the product site
- Starting offer
- Credits worth 3,000 decisions after email verification
Alternativas a Nimble
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads inside Firecracker micro-VMs. It is designed for workflows that need machine-level isolation, private networking between sandboxes, and programmatic control through SDK, CLI, or MCP.
GitBot
GitBot is a local browser-based workspace for creating reusable coding bots powered by Claude Code, Codex, or OpenCode. It helps developers run those agents across repository folders with persistent threads and configurable permissions.
AakarDev AI
AakarDev AI helps teams manage AI provider access, project-level setups, logs, and analytics from one dashboard. It supports BYOK workflows and lists providers including OpenAI, Google Gemini, Anthropic, Groq, Mistral AI, and Perplexity AI.
Trigger.dev chat agent
Trigger.dev chat agent is a durable AI chat backend for developers building stateful conversations that can survive refreshes, crashes, and long-running turns. It connects with the AI SDK `useChat` flow and runs on managed infrastructure with no timeout on a turn.
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repositories and verifies changes with the real compiler, debugger, sanitizers, and tests before showing a diff. It offers a free tier plus paid plans, with editor connectors and zero-retention handling described in the source.
Epherem
Epherem is a static security scanner for web apps built with AI coding tools such as Lovable, Bolt, v0, Replit, Cursor, and Claude Code. It checks source files and dependencies for common security mistakes, then provides categorized findings and suggested fixes without running the app.