Skip to main content
Manage Models is an independent feature for managing LLM provider metadata and per-model pricing. It is used across OpenLIT — including Chat, Evaluations, OpenGround, and Pricing — to look up model details and compute costs.

​
Overview

Every ClickHouse database connected to OpenLIT automatically gets a pre-populated set of 14 providers and 45+ models with default pricing. All models are fully editable — you can update pricing, add new models, or delete ones you don’t use.

​
Supported providers (out of the box)

ProviderExample Models
OpenAIGPT-4o, GPT-4o Mini, GPT-4 Turbo, GPT-3.5 Turbo
AnthropicClaude 3.5 Sonnet, Claude 3 Haiku, Claude 3 Sonnet
Google AIGemini 1.5 Pro, Gemini 1.5 Flash, Gemini 1.0 Pro
Mistral AIMistral Large, Mistral Medium, Mistral Small
GroqLlama 3.3 70B, Llama 3.1 8B, Mixtral 8x7B
PerplexitySonar Pro, Sonar, Sonar Reasoning
Azure OpenAIGPT-4o, GPT-4o Mini, GPT-4 Turbo, GPT-3.5 Turbo
CohereCommand R+, Command R, Command, Command Light
Together AILlama 3.1 70B Turbo, Llama 3.1 8B Turbo, Mixtral 8x7B
Fireworks AILlama 3.1 70B, Llama 3.1 8B, Mixtral 8x7B
DeepSeekDeepSeek Chat, DeepSeek Coder
xAIGrok Beta, Grok 2, Grok 3, Grok 3 Mini
Hugging FaceLlama 3 70B, Mistral 7B, Gemma 2 9B
ReplicateLlama 3 70B, Llama 3 8B, Mixtral 8x7B

​
Get started

1

Open Manage Models

Navigate to Manage Models in the sidebar under the Configuration section.
2

Browse providers and models

The left sidebar lists all providers. Click a provider to expand its models. Each model shows its context window and input price.
  • Default badge — a model that was pre-populated by OpenLIT.
  • Custom badge — a model you added manually.
3

Edit a model

Click any model to open the editor panel on the right. You can change:
  • Display Name — the human-readable name shown in dropdowns.
  • Model Type — chat, embeddings, images, or audio.
  • Context Window — the maximum token context length.
  • Input / Output Price per 1M Tokens — used by the Pricing feature to compute costs.
  • Capabilities — comma-separated list (e.g. function-calling, vision, streaming).
4

Add a new model

Click the + button next to any provider header to add a new model. Fill in the Model ID (the identifier your SDK sends, e.g. gpt-4o-2024-08-06), display name, and pricing.
5

Delete a model

Open a model in the editor and click the Delete button. Deleted models can be re-added at any time.

​
Export pricing JSON

GET /api/openground/models/export returns all models in an SDK-compatible pricing.json format, grouped by model type. This is useful for feeding custom pricing into the OpenLIT Python or TypeScript SDKs when the default pricing data doesn’t include your models.
Example response
{
    "chat": {
        "gpt-4o": {
            "promptPrice": 0.0025,
            "completionPrice": 0.01
        },
        "claude-3-5-sonnet-20240620": {
            "promptPrice": 0.003,
            "completionPrice": 0.015
        }
    },
    "embeddings": {
        "text-embedding-3-small": 0.00002
    }
}
Click the Export Pricing button in the top-right corner of the Manage Models page to download this file directly.

​
Public SDK Pricing URL

Each database config has a public, unauthenticated pricing URL that you can pass directly to the OpenLIT SDK. This lets your SDK load pricing from the same models you manage in the UI — no manual JSON files needed. The URL is shown in a bar below the header on the Manage Models page. Click Copy to copy it to your clipboard.
Click the SDK Usage button in the header to open a dialog with ready-to-paste code snippets for Python, TypeScript, and Go. Each snippet pre-fills the public URL and includes a copy button.
import openlit

openlit.init(
    otlp_endpoint="http://127.0.0.1:4318",
    pricing_json="http://your-openlit-host/api/pricing/export/<your-db-config-id>",
)
Endpoint: GET /api/pricing/export/{dbConfigId}
  • No authentication required — safe to embed in application configs.
  • Returns the same SDK-compatible pricing JSON format as the export button.
  • Responses are cached for 5 minutes (Cache-Control: public, max-age=300).
The URL contains your database config ID. While the endpoint only returns pricing data (no secrets or traces), share it only with systems that should access your model pricing.
The SDK fetches this URL on startup. If you update a model’s price in Manage Models, restart your app (or wait for the SDK’s refresh interval) to pick up the new pricing.

​
Import pricing JSON

You can bulk-import models from a JSON file. This is useful for syncing pricing across environments or importing from a colleague’s export.
1

Open the import dialog

Click the Import Pricing JSON button in the header of the Manage Models page.
2

Paste JSON

Paste either the structured format or the SDK pricing_json format.Structured format (recommended — includes provider and metadata):
{
  "models": [
    {
      "provider": "openai",
      "model_id": "gpt-4o-2024-08-06",
      "displayName": "GPT-4o (Aug 2024)",
      "inputPricePerMToken": 2.5,
      "outputPricePerMToken": 10
    }
  ]
}
SDK pricing_json format (provider defaults to "unknown" — edit after import):
{
  "chat": {
    "gpt-4o": { "promptPrice": 0.0025, "completionPrice": 0.01 }
  }
}
3

Import

Click Import Pricing JSON. Models that already exist (same provider + model_id) are automatically skipped — no duplicates are created.
Endpoint: POST /api/openground/models/import (authenticated)

​
Architecture

  • ClickHouse tables: openlit_providers (API key configs) and openlit_provider_models (all models with pricing).
  • No database_config_id column — each ClickHouse instance is its own scope. When you connect a new ClickHouse database, the migration automatically creates the tables and seeds all default models.
  • Platform library: src/lib/platform/providers/ — provider-registry.ts, models-service.ts, config.ts, default-models.ts.
  • API routes: GET/POST/DELETE /api/openground/models, GET /api/openground/models/export, GET /api/openground/providers, POST /api/openground/models/import, GET /api/pricing/export/{dbConfigId} (public).