小米 MiMo

This is one of the current flagship text models in Token Plan. The official subscription page explicitly includes it and lists its Credits deductions at 2.5 for cache hits, 300 for uncached input, and 600 for output, making it better suited to higher-value, more demanding coding and agent tasks.

This is the more everyday mainline text model inside Token Plan. The official page lists deductions of 2 Credits for cache hits, 100 for uncached input, and 200 for output, making it a better fit than Pro for most routine coding Q&A, code edits, documentation work, and sustained development sessions.

ASR is an explicitly supported speech-recognition model in the subscription system. The docs state that it does not consume Credits like text tokens but instead deducts by audio duration at 30M Credits per hour, which means Token Plan extends beyond text-only workflows into transcription, meeting notes, and voice-input development flows.

VoiceClone is one of the TTS-series models explicitly listed in the official docs and is currently marked as temporarily free in Token Plan without consuming Credits. It is better suited to scenarios where text output needs to become personalized voice feedback, assistant narration, or speech closer to a real speaker identity.

VoiceDesign is also part of the currently supported TTS family and is marked as temporarily free on the subscription page alongside the other TTS capabilities. Compared with plain text output, it is more suitable for voice-style design, narrated product experiences, and validating speech-oriented product prototypes.