> ## Documentation Index
> Fetch the complete documentation index at: https://docs.coplay.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Prompt Cost Guide

> Average prompt costs for Coplay users across GPT-5, Claude 4.5, and Gemini 3.

Coplay supports **Bring Your Own Key (BYOK)**. The data below represents the average cost of a single interaction (prompt + response) based on real-world usage metrics.

***

## 🏷️ Live Price List (Avg. Cost Per Prompt)

### 🟢 OpenAI Models

The industry standard for professional task reliability and tool-calling precision.

| Model                 | Avg. Cost / Prompt | Best Use Case                                         |
| :-------------------- | :----------------- | :---------------------------------------------------- |
| **GPT-5.2**           | **\$0.032**        | **The Expert.** High-stakes professional logic.       |
| **GPT-5.1**           | **\$0.031**        | **The Workhorse.** High-throughput reasoning.         |
| **GPT-5.1 Codex**     | **\$0.030**        | **Logic King.** Best for complex C# architecture.     |
| **GPT-5.1 Codex Max** | **\$0.023**        | **Deep Context.** Large file processing/scanning.     |
| **GPT-4.1**           | **\$0.035**        | **Legacy Logic.** Stable but higher avg. cost/prompt. |

### 🟣 Anthropic (Claude)

The leader in coding capabilities, particularly for "agentic" and multi-file workflows.

| Model                 | Avg. Cost / Prompt | Best Use Case                                                |
| :-------------------- | :----------------- | :----------------------------------------------------------- |
| **Claude 4.5 Opus**   | **\$0.110**        | **The Architect.** Unrivaled multi-file refactors.           |
| **Claude 4.5 Sonnet** | **\$0.100**        | **The Gold Standard.** Favorite for Unity devs.              |
| **Claude 4.5 Haiku**  | **\$0.025**        | **Speed Specialist.** Fast edits and sub-tasks.              |
| **Claude 3.7 Sonnet** | **\$0.026**        | **Hybrid Reasoning.** Produced in standard or thinking mode. |
| **Claude 3.5 Sonnet** | **\$0.082**        | **Legacy.** Maintained for project consistency.              |

### 🔵 Google & Specialized Models

Gemini 3 Pro is optimized for massive 1M+ token windows.

| Model               | Avg. Cost / Prompt | Best Use Case                                   |
| :------------------ | :----------------- | :---------------------------------------------- |
| **Gemini 3 Pro**    | **\$0.140**        | **Context Beast.** Massive codebase ingestion.  |
| **Gemini 2.5 Pro**  | **\$0.062**        | **Large Memory.** Reliable wide-context tasks.  |
| **Nova Premier VL** | **\$0.083**        | **Multimodal.** Advanced video/image logic.     |
| **Jamba Large 1.7** | **\$0.065**        | **High Throughput.** Optimized for low-latency. |

***

## 💰 Critical Billing Rules

<warning>
  **The 5-Minute Cache Window:** Anthropic and Google prices drop significantly when context is cached. However, if you wait more than **5 minutes** between replies, the cache expires, and you are billed the full price for your history again.
</warning>

### 📉 Efficiency Strategies

<Steps>
  <Step title="Prune Your Pins">
    **Unpin files** (`@`) as soon as you are done with them. Every pinned file adds to the input cost of *every* message in that thread.
  </Step>

  <Step title="The 200k Token Reset">
    Once a thread exceeds **200,000 tokens**, providers often implement a surcharge. When you hit this threshold, **Start a New Thread**.
  </Step>

  <Step title="Smart Routing">
    Use **Gemini 3 Pro** for deep codebase research, but switch to **Claude 4.5 Haiku** or **GPT-5.1** for writing individual methods to keep your costs low.
  </Step>
</Steps>