You can manually select the model when chatting with a character. Points vary based on the model, lorebook length, and response length. The figures below are estimates (est.) for typical requests, not a guaranteed number of turns. The maximum usage is reserved upon sending and settled based on actual usage after completion. Requests that did not return accurate usage due to interruptions will show the reserved amount as the estimated usage.
Kimi K3est. 23 points
GLM 5.3est. 7 points
Kimi K3.1 Proest. 15 points
Qwen 3.8 27Best. 5 points
DeepSeek V4 Flashest. 2 points
DeepSeek V4 Proest. 8 points
GLM 5.3 Flashest. 2 points
DeepSeek V4.1 Flashest. 2 points
DeepSeek V4.1 Flash Speedest. 5 points
MiMo V2.6 Proest. 1 point
Qwen3.6 35Best. 3 points
Gemma 4 31Best. 1 point