StartLux-Decision · 27B Auto-ThinkFREE STARTLUX TESTING

Free StartLux decisions.
Think when needed.

A decision model trained by StartLux Labs for agents and workflows. Send a state and typed questions; receive choices, scores and probabilities. Auto-Think reasons when its first answer is uncertain.

Free testing in the default group. A valid Bamboo API key is required. No top-up is required.

POST /v1/decisions
export BAMBOO_API_KEY="YOUR_API_KEY"
curl --request POST 'https://api.bamboonode.dev/v1/decisions' \
  --header "Authorization: Bearer $BAMBOO_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "startlux-decision-27b-autothink",
  "state": "A customer was charged twice for the same order and requests a refund.",
  "questions": {
    "team": {
      "type": "choice",
      "instructions": "Which team should handle this ticket?",
      "criteria": {
        "billing": "Payments and refunds",
        "tech": "Technical errors"
      }
    },
    "urgent": {
      "type": "noul",
      "instructions": "Does the customer report an immediate safety emergency?"
    }
  },
  "think_mode": "auto"
}'

73.4

Decision Index 0.3

27B Auto-Think · public part

86.9%

DecisionPivot

27B Auto-Think · contrast pairs across 7,800 core items

#1

Same-size open models

On the public part of Decision Index 0.3, the family leads at each published size from 0.8B to 27B.

Official StartLux release results · October 2026. The DecisionPivot score is the highest among the systems evaluated in that release. Bamboo Relay currently serves the 27B Auto-Think model.

Typed decisions

Ask choice, score and yes/no probability questions. Use structured results directly in your workflow.

Thinking on demand

Answer in one pass first. Auto-Think adds reasoning when the initial answer is uncertain.

Built for agents

Route support tickets, review tool calls and check agent steps. Set a probability threshold for human review.

Multiple providers · One API · Pay as you go

One API key.
Every leading model.

Access models from OpenAI, Anthropic, Google and more with one API key. Use OpenAI-compatible endpoints for chat and native APIs for specialized models.

quickstart.py
from openai import OpenAI

client = OpenAI(
    base_url="https://api.bamboonode.dev/v1",
    api_key="bk-••••••••••••••••",
)

resp = client.chat.completions.create(
    model="claude-sonnet-4-5",
    messages=[{"role": "user", "content": "Hello"}],
)

Connected providers

OpenAIAnthropicGoogle GeminiDeepSeekMoonshotQwen

Automatic failover

Configured backup channels can retry requests when an upstream fails.

Transparent usage

Inspect and export the model, tokens, latency, and cost for every request.

Credits as a firewall

Usage billing plus key-level limits keep budgets under control even if a key leaks.

Build with every model

Create an account and make your first API request.

Get started