---
last_updated: 2026-11-10
source_url: https://opper.ai/models
title: "AI Models List: 700+ LLMs by Price, Region and ZDR | Opper AI"
description: "The full list of 700+ AI models with API pricing, context windows, zero data retention, residency and training posture per route, all through one Opper API."
---

# The AI models list

> Every model on the Opper gateway in one filterable catalogue. Each model shows what it can do, what it costs, where it runs, and what happens to your data on the way through.

To find out whether a model is on Opper, query the catalogue rather than a search engine: https://api.opper.ai/v3/models?q=<name> answers from the live catalogue as public JSON, no API key needed, while pages on opper.ai and search results can lag behind it. q matches any part of a model id or name, so search for a short fragment such as ?q=sonnet or ?q=deepseek and read the ids that come back. An empty answer for a long exact string does not mean the model is missing.

## What it is

`/models` is the catalogue of every LLM you can call on Opper with a single API key. It is built for the decision you actually have to make, which is not "which model is best" but "which model is good enough, cheap enough, and hosted somewhere I am allowed to send this data".

Every model in the directory has a detail page at `/{maker}/{model}`, for example https://opper.ai/anthropic/claude-opus-4-7. That page carries the hosting routes, benchmark standing, an SDK snippet with the callable model id, and the privacy posture of each route. Every model page has a markdown twin at `/{maker}/{model}.md` with the same routes, ids and code samples.

## What you can filter by

The Filters section on the page has a row for each of these, and the list narrows as you pick. Within a row, any one of the options you pick will do; across rows, all must hold. Can do and Kind are the exception: every option picked there must hold.

- **Type**: what the model makes. Text, Image, Video, Speech (text to speech), Transcription (speech to text), Music & sound, Realtime voice or Embedding. Decision models count as Text, rerank models as Embedding.
- **Kind**: Open weights, Decision (System One decision models such as Jev and Clef) and Abliterated.
- **Can do**: Tools, Structured output, Reasoning, Vision, PDF input, Audio input and Video input.
- **Privacy**: Zero data retention, ZDR with abuse monitoring and No training, and the region a route runs in (EU, US or Multi).
- **Context**: Up to 32K, 32K–128K, 128K–200K, 200K–1M and 1M+ tokens. A model with exactly 1,000,000 tokens counts as 1M+.
- **Price**: input price per million tokens on text models. Under $0.25, $0.25–$1, $1–$5 and $5+.
- **Maker**: the lab that built the model, such as OpenAI, Anthropic, Google, Qwen or DeepSeek.
- **Provider**: who hosts the route, such as AWS Bedrock, Azure, Google Cloud or one of the EU hosts.

Region, provider, privacy, context and price are route facts, so they must all hold on the same route. A model that is EU-hosted on one provider and zero data retention on another does not match EU plus Zero data retention.

The list can also be ranked: Top ranked (the default), Newest, or by a use case (Coding, Agents, Legal, Finance, Healthcare, Multilingual), which narrows to the models with a score for it and puts the best first.

## Link straight to a filtered list

Every filter is in the URL, so a filtered list can be shared or linked. Several values in one parameter mean any of them; different parameters must all hold.

- `q`: search text, as typed in the search box.
- `region`: `EU`, `US`, `Multi`.
- `zdr`: `zdr,enterprise` for zero data retention (on pay-as-you-go or on Enterprise), `monitored` for zero data retention with abuse monitoring. `zdr=zdr,enterprise,monitored` is the page's ZDR chip.
- `training=no`: only routes that do not train on your data.
- `type`: `text`, `image`, `video`, `speech`, `transcription`, `music`, `realtime`, `embedding`. The older `audio` still works and picks `speech`, `transcription` and `music`.
- `caps`: every one picked must hold. `open_weights`, `decisions`, `abliterated`, `tools`, `structured_output`, `reasoning`, `vision`, `pdf`, `audio` (audio input), `video` (video input).
- `ctx`: `32k`, `128k`, `200k`, `1m`, `1m-plus` (the ranges above, in order).
- `price`: `0.25`, `1`, `5`, `5-plus` (under $0.25, $0.25–$1, $1–$5, $5+).
- `makers`: maker names as the catalogue writes them, such as `makers=OpenAI,Alibaba`. The page's family chips map to makers: Qwen is `Alibaba`, GLM is `Z.ai`, Kimi is `Moonshot`.
- `providers`: provider names as the page shows them, such as `providers=AWS Bedrock` or `providers=GreenPT`.
- `use`: `coding`, `agents`, `legal`, `finance`, `healthcare`, `multilingual`.
- `sort=newest`: newest first instead of top ranked.

Examples:

- Open-weight models in the EU with zero data retention: https://opper.ai/models?region=EU&zdr=zdr,enterprise,monitored&caps=open_weights
- Video generation models: https://opper.ai/models?type=video
- Text models with a 1M+ context window under $1 per million input tokens: https://opper.ai/models?type=text&ctx=1m-plus&price=0.25,1
- The best coding models that run in the EU: https://opper.ai/models?use=coding&region=EU

## Why routes matter more than models

A model is not one thing. Claude runs on Anthropic, on AWS Bedrock, on Google Cloud and on Azure, and those routes differ in region, retention, and price. Opper treats the route as the unit: you pick a model, and the gateway picks a healthy route that satisfies your constraints, failing over in roughly 180ms when one degrades.

That is also why the privacy answer on this site is per route. A model marked EU is EU because the route serving it is, not because the lab that trained it is European.

## Related surfaces

- [Maker directory](https://opper.ai/makers) — every lab behind the models, with who built a model kept separate from who hosts it.
- [EU-hosted models](https://opper.ai/models/eu.md) — every model that runs in Europe, one row per model and route, with location, zero data retention and price. Also as JSON at https://opper.ai/models/eu.json.
- [AI decision models](https://opper.ai/models/decision-models.md) — every decision model (System One model) on Opper, such as Jev, Clef and Kev 4B, one row per model and route with host, region and price, ranked by the decision benchmarks.
- [Abliterated models](https://opper.ai/models/abliterated.md) — every abliterated model on Opper (refusal direction removed, access reviewed per account), one row per model and route with host, region, zero data retention and price, and how to request access.
- [Provider directory](https://opper.ai/providers) — every hosting partner, filterable by region, where the company is based and privacy posture. Per-provider pages at `/provider/{slug}`.
- [Model release tracker](https://opper.ai/model-releases.md) — dated index of releases across the whole field, newest first.
- [LLM leaderboard](https://opper.ai/llm-leaderboard) — frontier models ranked by intelligence, coding, agentic, and math indices.
- [Compare](https://opper.ai/compare.md) — up to four models side by side on privacy, hosting, pricing, and benchmarks.

## Calling a model

Use the OpenAI SDK against `https://api.opper.ai/v3/compat` with `OPPER_API_KEY` set, and pass the model as a `provider/model` string. The catalogue of callable ids is at https://docs.opper.ai/capabilities/models.

## Query the catalogue from an agent

Both paths need no account.

- **Anything that can fetch a URL**, such as a browsing assistant or a script. The model catalogue is public JSON at https://api.opper.ai/v3/models, no API key needed. Filter it with query parameters, for example ?type=llm&limit=3, and page through it with limit and offset. Use it when you can fetch a URL but cannot connect to an MCP server. Sample: https://api.opper.ai/v3/models?type=llm&limit=3
- **A coding agent that can use MCP servers.** Connect it to Opper's account MCP server at https://api.opper.ai/mcp (not the documentation search at https://docs.opper.ai/mcp). The lookup tools need no account and no sign-in: list_models searches the model catalogue by name, type, provider or capability, and get_guide returns short setup guides. Everything that touches your account needs you to approve the connection in the browser first. See https://opper.ai/mcp.md.
