One question. About a third of a cent.
That's a typical exchange, all in. No subscription, no minimum. You hold a balance; each message draws only what it uses, at the rate of whichever model you pick.
Illustrative · a typical glm-4.7 exchange
What’s on the network.
Open-weight and frontier models, listed with their capabilities and per-token rates. A million tokens is roughly 750,000 words.
glm-4.7-flashglm-4.7-flashfreeTools128kFreeFree
openai/gpt-oss-120bgpt-oss-120bReasoningTools128k$0.05$0.29
deepseek-ai/DeepSeek-V4-FlashDeepSeek-V4-FlashReasoning128k$0.12$0.24
meta-llama/llama-4-maverickllama-4-maverickVisionTools1M$0.18$0.60
qwen/qwen3-coderqwen3-coderToolsCode256k$0.20$0.80
xai/grok-4-fastgrok-4-fastReasoningTools2M$0.20$0.50
google/gemini-3-flashgemini-3-flashVisionImage toolsTools1M$0.30$2.50
minimax/minimax-m2minimax-m2Tools200k$0.30$1.20
mistralai/mistral-large-3mistral-large-3Tools128k$0.60$1.80glm-4.7glm-4.7ReasoningToolsCode200k$0.73$2.66
anthropic/claude-sonnet-5claude-sonnet-5VisionReasoningTools1M$2.66$13.31
black-forest-labs/flux-2flux-2GenerationEdit—$0.03 / imageNo tiers. No fine print.
The number in the table is the number you pay. Here’s what’s behind it.
Run a model yourself →Keep your twenty dollars.
Heavy use costs more; the meter only runs while you’re asking. For most people, that’s the cheaper deal, and the private one.
Charged whether you asked one question or a thousand. Tied to an account, a card, and a history of everything you said.
About ninety exchanges on glm-4.7. Nothing on file, nothing renewing, and a balance you can withdraw.
Try it before you fund it.
Your first questions are free. Then add $5 and watch what it actually costs.