One endpoint.
Point your SDK at Eco.
Keep your OpenAI-compatible client and your Orbio key. Just change the base URL.
Eco
Spend Orbio
credits smarter.
Route auto to the cheapest model that still fits.
Scroll // Reassemble
Break it down.
Route it smarter.
One request. The cheapest model that still fits.
A handcrafted CRT monitor separates into its casing, tube, bezel, and stand. The pieces follow a low-cost routing lane, then reassemble. The screen illustrates an everyday chat request using auto/balanced routing to DeepSeek V4 Flash. This is a routing illustration, not a live request.
01 / The shorter route
Big models for big problems.
Not every problem.
One small change.
Let auto do the choosing.
Use auto/cheap, auto/balanced, or auto/best. Eco matches the task to an eligible model.
A lighter bill.
Pay for what the job needs.
Eco picks the lowest output price in the selected tier. Every route comes with a reason.
02 / Make your credits go further
Same credits.
More possibility.
What are you working on?
Short answers, quick rewrites, the everyday stuff.
Illustration for 1M input + 1M output tokens. Compares Claude Opus 5.5 at lab and Orbio prices with the cheapest model in the example tier. Actual routes vary; classifier costs are excluded.
lab$ list price
$24.00Claude Opus 5.5
orbio$ same model
$18.00Same model · 25% less in seed pricing
eco$ routed estimate
DeepSeek V4 Flash
99.0% less than the Orbio baseline
deepseek/deepseek-v4-flashIN: 1M TOKENSOUT: 1M TOKENSUSD
Seed catalog · Price snapshot 2026-10-01 · USD estimates, not a quote
03 / A small diff. A smarter default.
Keep your stack.
Change the route.
Your SDK. Your Orbio key. Two environment variables and one model alias.
Try it in the playground# Your Eco endpoint + your own Orbio key
export OPENAI_BASE_URL="https://api.orbioeco.sh/v1"
export OPENAI_API_KEY="sk-orbio-…"
# In your existing chat completion request
model: 'auto/balanced'
Pin a model ID anytime to pass straight through.
04 / Good to know
A few details.
Do I need another API key?
No. Bring your own sk-orbio-… key. The playground saves it in localStorage on your device and sends it only to your configured Eco API, which forwards it to Orbio. Eco holds it in memory for that request and never persists it. You can remove the saved key anytime.
Will Eco change every request?
Only when you ask. Use auto, auto/cheap, auto/balanced, or auto/best to enable routing. Any other model ID passes straight through to Orbio without a classifier or model rewrite.
How does Eco choose a model?
Eco checks required capabilities, then uses deterministic rules to estimate difficulty. When confidence is low, a cheap classifier helps choose the tier. Eco picks the cheapest eligible model by output price within that tier. Preview and Send each route independently, so both may incur classifier cost and can choose different models.
Does Eco retain my prompts?
Eco never logs or stores prompts or completions. The API logs request metadata, and the playground saves only mode, selected model, and timestamp for recent routes. Prompts stay in the current page session. Your requests still go to Orbio and the model provider; their retention policies apply.
Is Eco part of Orbio?
Eco is complementary tooling for the Orbio ecosystem. Your credits and billing stay with Orbio. Eco helps you choose where to spend them; it does not sell credits or hold a balance.