Features

• Smart Splitting

  One request, several models, a smaller bill, decided for you on every call. When we can save, we save.

• Smart routing

  The right model and reasoning setting for every request, picked from evidence, with the reason on every response.

• Jev for classification

  Tagging, routing, moderation and sentiment for a fraction of the price of a chat model, named or picked by auto.

• Caching

  Your history cached for you, the 1 hour cache when turns are slow, and never a cut of your cache discount.

• Flex

  Half price when you can wait, with one deadline field.

• Batching

  Thousands of requests in one file at about half price, with auto routing on every line.

• Teams and keys

  Organizations, member quotas, keys with spend caps and model lists, and your own vendor keys with no service fee.

• Smart Splitting

  When a request carries a long document and asks about part of it, a cheap model copies out the lines that matter, Jev checks nothing is missing, and the model auto picked answers from the excerpt. In our tests it cut the cost of long-document questions by about a third, with every fact kept. When we can save, we save. When we can't, nothing changes.

• Organizations

  Create as many as you need, each with its own credits, keys, members, data and settings. One sign-in moves between them.

• Team members with quotas

  Invite people as owners, admins or members. Give each member a monthly budget in USD, and their keys stop when it is reached.

• Unlimited API keys

  Each key has its own monthly spend cap, request and token limits per minute, the models it may use, a fallback rule and a project for grouping.

• Your models, your call

  Switch any model or vendor off for the whole organization and auto never picks it. A key can narrow the list further.

• Prompt caching

  On by default. Zabio marks your system prompt, tools and conversation history for Anthropic's cache, so agent loops stop paying full price for their history on every turn. OpenAI caches on its own, slow conversations get a 1 hour cache automatically, and cached tokens are billed at the vendor's cached rate.

• Switch models, keep the conversation

  Stored conversations follow the thread across models and vendors. When the history outgrows the next model, Zabio summarizes the oldest part and carries on.

• Automatic failover

  When a provider fails or stalls, Zabio moves on to the next model. For models you name, choose same vendor, any vendor or no fallback, per key.

• Long calls that finish

  Reasoning models that think for minutes get up to 10 minutes per attempt, streamed or not. Zabio keeps the connection alive, so nothing in between gives up first.

• Your own vendor keys

  Plug in your own Anthropic, OpenAI or other vendor keys and pay no service fee on those calls; auto requests pay only the small fixed fee for automatic model choice. Routing, conversations and the dashboard keep working.

• Cost and saving on every call

  Each response carries its cost and its saving. The dashboard splits spend by model, key and member, shows where the tokens went (history, tool results, cache and reasoning), and exports every call to CSV.

• Private by default

  Prompts are not stored unless you turn on body logging or use stored conversations. Stored content is encrypted per organization, kept 1 to 365 days, and every read lands in the audit log.

• Web app, CLI and playground

  Manage everything from the dashboard or the zabio command line, and try any model in the playground with code you can copy and run.

• Typed decisions with Jev

  Tagging, routing, moderation and sentiment for a fraction of the price of a chat model: name jev-1.13.0 with a JSON schema of fixed choices, yes or no and scores, and get calibrated answers in about 0.2 seconds, around 0.00002 USD a call. Through auto, Zabio uses it on its own whenever it is the cheaper answer.

• Half price when you can wait

  Add a deadline of 15 minutes or more to a request and Zabio serves it on the flex tier of OpenAI or Google where the model has one: the same model at about half the price, just slower. If the vendor has no capacity, it runs at the normal price instead, so it never fails because of flex.

New: batching. Send thousands of requests in one file and pay about half, with auto routing on every line.