On this page 5 sections

LLM token and cost estimate for one prompt across 18 chat models in VS Code

What the AI tools do

Tokens and cost

LLM Models, Tokens & Cost estimates the tokens in a prompt and its per-request, daily and monthly cost across 18 chat models, and compares current chat models' context windows, output limits and prices.

Prompt Builder

Fill a prompt template with your variables and check it for common mistakes before it reaches a model.

Output Cleaner & JSON Validator

Models wrap JSON in prose and code fences, and add trailing commas. The cleaner extracts the JSON, repairs it, validates it and tells you what it fixed.

Smaller prompts and starter code

JSON → TOON shrinks structured data for prompts. LLM Client Setup generates a client for OpenAI, Anthropic, Gemini, Azure, Ollama or any OpenAI-compatible server, and AI App Starter scaffolds a small project with config, a client with retries, a prompts folder and tests.

Call real models

The REST API client has request shapes for OpenAI, Anthropic, Gemini, Azure OpenAI and Ollama. Single prompts, streaming and schema validation are free (up to 25 AI requests, 25 prompt runs and 10 streamed responses a day); model comparison, benchmarks, embeddings and RAG pipeline tests use points.

Machine learning helpers

GPU memory and speed estimates, a model serving starter, model metrics, dataset split and learning-rate schedule planners, and a model card template.

Example: clean up a model reply

A reply like this cannot go straight into JSON.parse:

Model reply

Sure! Here is the result:
```json
{
  "name": "Ada",
  "tags": ["admin", "beta",],
}
```
Let me know if you need anything else.

The Output Cleaner returns valid JSON and lists its fixes: Extracted the fenced code block and Removed trailing commas.

JSON

{
  "name": "Ada",
  "tags": ["admin", "beta"]
}

Get started

Run DevSnip Pro: LLM Models, Tokens & Cost, or press Ctrl/Cmd+K in the Tools sidebar and search for "tokens".
Paste your prompt.
Set requests per day to see daily and monthly cost, and compare models.
Run DevSnip Pro: Browse AI & ML Tools to see the rest of the section.

TipToken counts and prices are estimates for planning. Your provider's tokenizer and current price list are the final word on billing.

Tools in this area

AI & ML 14 tools

Prompt Builder
LLM Models / Tokens & Cost
LLM Client Setup
LLM Output Cleaner & JSON Validator
LLM API Tester
JSON → TOON
AI App Starter
AI/ML Code Snippets
GPU Memory & Speed
Model Serving Starter
Model Metrics
Dataset Split Planner
Learning-Rate Schedule
Model Card

Search all the developer tools →

Questions

Do I need an API key?: Not for token counts, cost estimates, the Prompt Builder or the Output Cleaner: they run locally. You need a provider key only to call a real model from the REST API client.
Is my prompt sent anywhere?: No. The estimating and cleaning tools work offline. Of the toolbox tools, only the LLM API Tester uses the network, and it is marked "Uses network".

LLM token counts, costs and AI tools in VS Code