Models catalogue | EvaEsi Dev

Endpoint

GET https://evaesi.dev/api/v1/models

Selection

Omit model on /api/v1/responses and the server uses instant by default. Pass an explicit identifier from the table below to choose another model.

The catalogue

Each family ships several versions. They share a backend and differ in how hard the model works, how much context it carries, and what it costs. Rates below are read live from the published card.

id	Name	Input / M	Output / M	Max in	Max out
`instant`	EvaEsi Instant	£0.09	£0.30	100000	32000
`instant.fast`	EvaEsi Instant Fast	£0.14	£0.45	100000	16000
`reasoning`	EvaEsi Reasoning	£0.36	£1.20	100000	64000
`reasoning.deep`	EvaEsi Reasoning Deep	£0.42	£1.40	200000	96000
`reasoning.verify`	EvaEsi Reasoning Verify	£0.45	£1.50	100000	32000
`reasoning.fast`	EvaEsi Reasoning Fast	£0.54	£1.80	100000	32000
`elements`	Elements	£0.80	£2.70	100000	64000
`elements.fast`	Elements Fast	£1.00	£3.30	100000	32000
`elements.codebase`	Elements Codebase	£1.15	£3.75	400000	64000

Every version shares a 4096-byte context window and supports streaming. Batch and tool-call surfaces are not yet exposed on any of them.

The catalog response also includes evaexplore-1 (preview: true, available: false, no context window). It is not callable. Filter on available before use.

Older model names

Older or human-readable identifiers normalise to the canonical IDs above. Either form is accepted in POST /api/v1/responses.

Alias	Resolves to
`eva`, `eva-1`, `EvaEsi`, `evaesi-v1`, `evaesi-instant`, `evaesi_instant`, `esi-standard`, `eva-standard`, `conversation`, `creative`	`instant`
`evaesi-reasoning`, `evaesi_reasoning`, `analytics`, `reason`	`reasoning`
Hyphenated version ids, e.g. `elements-codebase`, `reasoning-deep`	The dotted form, e.g. `elements.codebase`
`esielement`, `esielement-1`, `element`, `code`, `coding`	`elements`

The response

{
  "models": [
    {
      "id": "instant",
      "name": "EvaEsi Instant",
      "description": "Fast general chat with live evidence.",
      "capabilities": ["text", "conversation", "analysis"],
      "max_input_rungs": 100000,
      "max_output_rungs": 32000,
      "available": true,
      "supports_streaming": true,
      "supports_batch": false,
      "supports_tools": false,
      "context_window": 4096,
      "preview": false,
      "deprecated": false,
      "sunset_at": null
    }
  ]
}

preview is true while a model's API surface may still change without notice.

Example

cURL

curl https://evaesi.dev/api/v1/models \
    -H "Authorization: Bearer esi_dev_YOUR_KEY"

Python

import requests

response = requests.get(
    "https://evaesi.dev/api/v1/models",
    headers={"Authorization": "Bearer esi_dev_YOUR_KEY"},
)

for model in response.json()["models"]:
    status = "available" if model["available"] else "not available"
    print(f"{model['id']}: {model['name']} ({status})")

JavaScript

const response = await fetch("https://evaesi.dev/api/v1/models", {
  headers: { "Authorization": "Bearer esi_dev_YOUR_KEY" },
});

const { models } = await response.json();
models.forEach(model => {
  console.log(`${model.id}: ${model.name}`);
});

Start from a live reference page.