Skip to content

Cost Dashboard

The Cost Dashboard tracks token usage across all your sessions and estimates what the same workload would cost on commercial cloud APIs.

Sidebar item: Costs Route: /costs

Why Track Costs Locally?

Ollama runs on your machine at zero cost. But understanding the equivalent cloud cost helps you:

  • Quantify savings — See how much you're saving by running locally
  • Plan capacity — Understand which models consume the most tokens
  • Compare efficiency — Identify which models give the best output per token

Dashboard Sections

Summary Cards

Four cards at the top showing:

  • Total Tokens — Sum of all prompt + completion tokens across sessions
  • Sessions — Total number of chat sessions recorded
  • Estimated Cost — What the same usage would cost on cloud APIs
  • Avg Cost/Session — Average estimated cost per session

Token Usage by Model

A horizontal stacked bar chart showing prompt (indigo) and completion (purple) tokens for each model you've used. Models are sorted by total token count.

Daily Usage

A line chart with two axes:

  • Left axis — Total tokens per day (filled area)
  • Right axis — Estimated cost per day (dashed line)

Hover over any point to see exact values.

Model Breakdown Table

A detailed table with columns:

ColumnDescription
ModelThe Ollama model name
SessionsNumber of chat sessions
PromptTotal prompt tokens
CompletionTotal completion tokens
TotalCombined token count
Est. CostEstimated cloud-equivalent cost
SourceWhere the pricing data comes from

Pricing Data

Pricing is based on equivalent cloud API rates for common model families:

Model FamilyInput $/1MOutput $/1MSource
llama3.2$0.04$0.04Together AI
llama3.1$0.05$0.08Groq
gemma3$0.10$0.10Google
mistral$0.25$0.25Mistral
deepseek-r1$0.55$2.19DeepSeek
phi4$0.07$0.14Azure
qwen2.5$0.15$0.15Alibaba

Models not in the pricing table fall back to a generic estimate of $0.10/$0.10 per 1M tokens, labeled as "Estimated".

These are estimates

All costs shown are what equivalent usage would cost on cloud APIs. Ollama runs locally at no cost. The pricing data is a static snapshot and may not reflect current cloud pricing.

Data Source

The cost dashboard derives all data from the metrics store — the same session metrics recorded during chat. No additional tracking or storage is needed. The cost store is purely computed (reactive) and recalculates automatically when new sessions are recorded.

Released under the Apache 2.0 License.