> ## Documentation Index
> Fetch the complete documentation index at: https://docs.runchat.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Credits
> Understand how Runchat credits work, how they're consumed, and current model pricing
Credits are Runchat's universal currency for running AI models and tools across text, image, video, and 3D generation.
## How credits work
When you run a model, Runchat calculates the credit cost based on the model's unit pricing. Different models charge in different units:
* **Token-based models** (LLMs), charged per token for input and output separately
* **Image models**, charged per image generated
* **Video models**, charged per video or per second of video
* **3D models**, charged per mesh generated
## Buying credits
Every account starts with **500 free credits** on the Starter plan. You can get more credits by:
* **Credit top-ups**, pay-as-you-go credits, \$1 per 1000 credits
* **Subscribing to a plan**, monthly credits included with Hobby (\$15 of credits/mo), Pro (\$50 of credits/mo), and Max (\$150 of credits/mo) plans
Unused credits roll over on paid plans: plan credits and credit top-ups each carry forward for **3 months** after the billing cycle that granted them ends. Starter plan credits don't roll over, but top-ups purchased on Starter keep the same 3-month window. You can see exactly when rolled-over credits expire on your usage dashboard.
## Free models
Hobby, Pro and Max plans all include access to free models for chat and image generation, plus web search. These let you prototype and explore ideas without worrying about the cost of each generation or lookup.
Free access is subject to daily, weekly and monthly caps that increase with the Pro and Max plans. Web search draws from the same free pool, so it won't cost credits until that pool is exhausted.
If you exceed your free cap you can upgrade your plan, wait until the next day
or use Runchat credits instead.
## Bring your own keys
On the **Hobby** plan and above, you can link your own API keys from providers like OpenRouter, Fal, Gemini, and Groq. When using your own keys, model usage is billed directly to your provider account instead of consuming Runchat credits. See [Credentials](/concepts/integrations/credentials) for setup.
## Fixed credit costs
Some actions have a fixed credit cost on top of any model cost:
| Action | Credits |
| -------------- | ------- |
| Running a node | 1 |
| Running an app | 1 |
| Web search | 10\* |
\* On Hobby, Pro and Max plans, web search draws from the [free pool](#free-models) and only costs credits once that pool is exhausted.
Apps are Code Nodes that run on their own compute and serve their own UI. Any
request to these apps will be billed as one credit.
## Cost warnings
Before you run a node, you'll see an estimated credit cost on the run button. When a run is going to cost more than 1,000 credits (roughly \$1 on the Starter plan), the bolt icon turns orange. The same orange badge appears on the selection toolbar when running multiple nodes at once.
The estimate is approximate. The actual cost is deducted after the run completes based on what was really used.
## Model pricing
The tables below show the current credit cost for all available default models. Prices are listed per unit (per token, per image, etc.).
For **token-based models**, the cost depends on the number of input and output tokens. Models with prompt caching offer reduced rates for cached input tokens.
You can install additional models from [OpenRouter](https://openrouter.ai) or
[fal.ai](https://fal.ai) by pasting a model URL into the Search bar in the
Editor.
### Text models (per token)
| Model | Output | Input | Cached Input |
| -------------- | ------ | ------ | ------------ |
| Claude Opus 5 | 0.05 | 0.01 | 0.001 |
| Gemini 3 Flash | 0.008 | 0.001 | 0.0002 |
| Gemma 4 31B | 0.003 | 0.002 | — |
| GPT-5.6 Luna | 0.002 | 0.0004 | 0.0000 |
| Kimi K3 | 0.03 | 0.006 | 0.001 |
| Mimo 2.5 | 0.0006 | 0.0003 | 0.0000 |
### Image models
| Model | Price | Unit |
| ------------------------------ | ----- | ----------------------- |
| Flux 2 Pro | 100 | per image |
| Topaz Upscale | 300 | per image |
| Crisp Upscale (Recraft) | 12 | per image |
| Extract Video Frame | 12 | per image |
| FLUX 2 Lora Edit | 42 | per processed megapixel |
| Flux LoRA | 80 | per megapixel |
| FLUX.2 \[klein] 9B LoRA | 40 | per megapixel |
| GPT Image 2 | 40 | per image |
| Ideogram V4 | 15 | per megapixel |
| Ideogram V4 Edit | 30 | per megapixel |
| Ideogram V4 Tiling | 120 | per megapixel |
| Imagen 4 Fast | 40 | per image |
| Imagen 4 Pro | 80 | per image |
| Mask Image (EVF) | 12 | per image |
| Nano Banana (Gemini 2.5) | 80 | per image |
| Nano Banana 2 (Gemini 3.1) | 160 | per image |
| Nano Banana 2 Lite | 80 | per image |
| Nano Banana Pro (Gemini 3) | 300 | per image |
| Nucleus Image | 20 | per megapixel |
| Qwen Image 2 | 72 | per image |
| Relighting | 80 | per image |
| Remove Background (Bria) | 52 | per image |
| Runchat Image | 4 | per image |
| Seedream 5.0 Pro Image Editing | 300 | per image |
| SeedVR2 | 2 | per megapixel |
| SVG to PNG | 6 | per image |
| Vectorize (Recraft) | 20 | per image |
| Z-Image Turbo Seamless Tiling | 40 | per megapixel |
### Video models
| Model | Price | Unit |
| ---------------------------------------- | ----- | -------------- |
| Combine Videos | 6 | per generation |
| Kling 2.1 (pro) | 200 | per second |
| Kling 2.5 | 100 | per second |
| Kling O1 First Frame Last Frame to Video | 200 | per second |
| MiniMax Hailuo 02 | 100 | per second |
| Seedance 2.0 Fast Image to Video | 480 | per second |
| Veo3.1 Lite FLF | 100 | per second |
| Veo3.1 Lite Image to Video | 100 | per second |
### 3D models
| Model | Price | Unit |
| ---------------------------- | ----- | -------------- |
| Hunyuan3d V3 | 800 | per mesh |
| Trellis 2 | 600 | per mesh |
| Hunyuan 3D Rapid Image to 3D | 500 | per mesh |
| Marble Mini | 400 | per generation |
| Pixal3d | 600 | per generation |
| Trellis | 40 | per mesh |
| Trellis 2 Fast | 80 | per generation |
| Tripo H3.1 Image to 3D | 400 | per generation |
| Triposplat | 100 | per generation |
### Audio models
| Model | Price | Unit |
| ------------------------- | ----- | ---------- |
| Sonilo V1.1 Text to Music | 5 | per second |
## Next steps
* [Subscriptions](/concepts/teams/subscriptions): plan tiers, included credits, and team seats
* [Credentials](/concepts/integrations/credentials): link your own provider API keys to bypass Runchat credits
The estimate is approximate. The actual cost is deducted after the run completes based on what was really used.
## Model pricing
The tables below show the current credit cost for all available default models. Prices are listed per unit (per token, per image, etc.).
For **token-based models**, the cost depends on the number of input and output tokens. Models with prompt caching offer reduced rates for cached input tokens.