Developers

MCP server

Connect the index to Claude, Cursor or Codex and ask about prices without leaving your editor.

Endpoint (Streamable HTTP)

https://mcp.inferindex.dev/mcp

Connect

Terminal

claude mcp add --transport http inferindex https://mcp.inferindex.dev/mcp

~/.cursor/mcp.json

{
  "mcpServers": {
    "inferindex": {
      "url": "https://mcp.inferindex.dev/mcp"
    }
  }
}

Terminal

codex mcp add inferindex --url https://mcp.inferindex.dev/mcp

~/.codex/config.toml

[mcp_servers.inferindex]
url = "https://mcp.inferindex.dev/mcp"

Cheapest provider for DeepSeek V3.2 that doesn't train on prompts?

Tool call

cheapest({"model":"deepseek/deepseek-v3.2","no_training":true,"strict":true})

DeepInfra, $0.29 per 1M tokens blended ($0.26 input, $0.38 output).

The answer is the API's own result while this page was generated.

search_models(query)
Search the catalogue
“Which DeepSeek models do you track?”
cheapest(model, no_training, region, strict, prompt_tokens, limit)
Cheapest offer for a model, with its conditions
“Cheapest provider for Llama 3.3 70B that doesn't train on prompts?”
compare_providers(model, sort, region, include_tiers)
Every provider for a model, side by side
“Compare every provider of Qwen3 235B A22B 2507, EU region only.”
price_history(model, days, at, granularity)
Price over a period or at a date, plus the lab's official prices
“How did the price of GPT-4o mini change since 2024?”
estimate_cost(model, prompt_tokens, output_tokens, requests_per_day)
Cost of a workload, per request and per month
“What would 20,000 requests a day of 2,000 input and 500 output tokens cost on DeepSeek V3.2?”
provider_ranking()
Providers ranked by their gap to the market price
“Which API providers are the cheapest overall?”

cheapest
```
What is the cheapest way to call DeepSeek V3.2 right now?
```
the price, the seller, the date it was checked, and a flag when the price is a promotion.

estimate_cost

I send about 2,000 tokens in and get 500 out, 1,000 times a day, on gpt-oss-120b. What will it cost per month?

cost per request and per month, with the assumptions repeated.

self_host_or_api
```
Is it cheaper to self-host Qwen3 32B than to use an API?
```
the break-even volume and the throughput needed as a condition; hardware rental only.