Shape Credits Guide

We’ve provided a chart to help give you insight to Premium Shapes and credit usage

Credit Usage Chart

Keeping track of your credit balance is essential for all Premium Shape users. Your credits are consumed at different rates depending on several factors.

The most significant factor is your chosen premium engine (or the one selected by your shape creator). While we don’t share the exact conversion formula, the guide below helps you understand how different models affect message efficiency and credit usage.

Premum AI Engines - Cost Efficiency Ranking

AI Engine	Message Efficiency vs Sonnet 4.6
Kimi K2 0905	~7.5x more messages than Sonnet 4.6
Kimi K2.5	~6.9x more messages than Sonnet 4.6
GLM 5	~6.0x more messages than Sonnet 4.6
Gemini 3 Flash Preview	~5.1x more messages than Sonnet 4.6
Hermes 4 405B	~4.5x more messages than Sonnet 4.6
ChatGPT-5.4 Mini	~3.4x more messages than Sonnet 4.6
Claude Haiku 4.5	~3.0x more messages than Sonnet 4.6
Gemini 2.5 Pro	~1.6x more messages than Sonnet 4.6
Gemini 3.1 Pro Preview	~1.3x more messages than Sonnet 4.6
ChatGPT-5.3 Chat	~1.1x more messages than Sonnet 4.6
ChatGPT-5.4	~1.0x (same as Sonnet 4.6)
Claude 3.7 Sonnet	1.0x (same as Sonnet 4.6)
Claude 3.7 Sonnet (Thinking)	1.0x (same as Sonnet 4.6)*
Claude Sonnet 4.6	1.0x (baseline)
Claude Opus 4.6	0.6x messages vs Sonnet 4.6
Claude 3.5 Sonnet	0.5x messages vs Sonnet 4.6

* Effective cost higher in practice due to thinking tokens.

The message efficiency calculations are based on a weighted average of input/output costs compared to Claude Sonnet 4.6’s baseline, assuming typical conversation patterns.

Offered premium engines may change over time and may not reflect the most up to date available engines. This chart is to serve as a general rule of reference. Actual experience may vary based on shape design and settings. All AI Engines are available at direct cost pricing, pricing may shift over time based on our provider’s pricing changes. Last updated: April 2026.

Tips on Credit Usage

Understanding Credit Usage

Every time your Shape sends a message, it talks to an AI engine that processes all the context you’ve given it — your messages, the Shape’s settings, and any memories or instructions you’ve enabled.
The more context = the more tokens used = the more credits spent.

What Counts Toward Input Tokens

Each Shape message includes multiple “inputs” that tell the AI who it is, what’s happening, and what to say.
Here’s what gets counted:

Input Source	What It Is	Token Impact
Your Message	What you just type or say each time you chat.	🟢 Low – Short messages barely cost anything; long paragraphs increase cost slightly.
Training Samples	Example chats you’ve added to “teach” your Shape. Managed on the AI Engine page.	🟢 Low – Only up to 10 examples are pulled at once, so cost stays small.
Knowledge	The Shape’s linked knowledge base or uploaded reference material. Controlled on the AI Engine page.	🟢 Low – Up to 10 knowledge entries may load per message depending on settings. Each one adds input length.
Long-Term Memory	Saved memories from past sessions. Found on the Shape’s Memory page and AI Engine settings.	🟢 Low – A maximum of 10 memories are pulled each time.
Shape Personality & Preset	The Shape’s core traits, examples, and writing style you set during creation.	🟡 Medium – Brief traits are fine, but long, detailed personalities add up.
Custom Shape Instructions	Extra behavior rules you set in the Shape’s configuration menu.	🟠 High – Costs scale with length. Keep them concise when possible.
Chat Instructions	Custom context that applies to everyone in a Chat. Found in Chat → AI Configuration.	🟠 High – The longer the instructions, the more each Shape message costs.
Your Persona	The personal profile or “who you are” context you’ve entered.	🟠 High – There’s no input limit here, so long personas can quickly consume tokens.
Short-Term Memory	The Shape’s recall of recent chat messages. Managed on the AI Engine page.	🔴 Very High – If set to 200, it re-reads all 200 recent messages each time. Great for remembering details, but expensive.

Settings That Affect Cost

You can control how much input each message sends by adjusting these toggles in your Shape or Chat:

Setting	Effect on Cost	Tips
Engine Choice	Different AI models have different prices per token.	💡 Cheaper engines = longer chats for less.
Memory Settings	Control how much context is re-sent.	💡 Try lowering if your Shape doesn’t need full recall.
Knowledge Settings	Enables use of stored data or uploaded text.	💡 Turn off if not needed for casual chats. Utilize the Engine’s built in data.
Training Level	Controls how many examples are sent to guide tone/style.	💡 Keep to Low unless you need better replication.
Chat Settings	Custom instructions, Shape instructions, Persona	💡 You can be very concise with all of your inputs here!

---

Example Scenarios

Example	What’s Happening	Credit Impact
“Casual chat with a short prompt”	You send short messages, low memory, no knowledge.	🟢 Very low.
“Storytelling with detailed Shape personality”	Long custom prompt + long Shape responses.	🟠 Moderate.
“Research Shape with knowledge docs + training”	Pulls from multiple sources + high memory.	🔴 High.
“High-end engine with high short-term memory ON”	Expensive model + large input context each turn.	🔴🔴 Very high.

Tips to Stretch Your Credits

• Keep inputs concise. Every word adds tokens.
• Use memory settings wisely unless you need deep recall.
• Disable unused features like knowledge or training.
• Switch engines if you’re chatting often, try a different one.

TL;DR

Your Shape’s total token input =
(Your message) + (Shape settings) + (Memories) + (Instructions) + (Knowledge/Training)
More context = richer replies, but also more credit use.

Return to App