MiniTavern v1.5.0 & v1.5.1: LLM config, member models, and chat UX
MiniTavern v1.5.0 and the follow-up v1.5.1 emergency patch are now live on both the App Store (iOS) and Google Play (Android). You can download or update from the stores directly.
The main changes across both releases focus on LLM configuration, member models, and chat interaction.
Recent changes
Recent changes include:
- LLM configuration — OpenAI request format, KoboldCpp, LM Studio, Kimi, OpenRouter
- Exclusive / PRO member models — GPT-5.5, Claude Fable 5, long-term model maintenance
- Global model config — one app-wide selection instead of per-character-card model middle layer
- Chat UX — quick model switch, model status copy, message bottom actions, LLM prompt cards
- Local conversation models — confirm before use, stop while a request is in flight
- Character card import — unified entry point
- Stability — selected-model persistence, OpenRouter add flow, framework crash fixes on specific devices
What changed (v1.5.0 + v1.5.1)
LLM configuration
- Broader general request formats, including an OpenAI-compatible path that fits local stacks such as KoboldCpp and LM Studio.
- Dedicated Kimi configuration entry.
- OpenRouter add flow fixed after a protocol update (v1.5.1).
- New provider entry points on the LLM page are collapsed by default, so existing configs stay easier to manage.
Member models and credits
- Exclusive member models keep receiving maintenance and upgrades; current highlights include GPT-5.5 and Claude Fable 5, while basic models remain for everyday use.
- Stronger credit granting for PRO membership categories.
- Profile display when credits are 0 is cleaner; custom LLM setups no longer show a misleading “No request credits available” state.
Chat and model switching
- Quick model switching on the conversation page is easier when a model faults or underperforms.
- The per-character-card model configuration middle layer is removed; use a single global model configuration.
- Selected LLM persistence is fixed: switching a model on the conversation page and returning to the list now keeps the choice (v1.5.1).
- Local conversation models get a confirmation prompt, support stop while sending, and cleaner prompt-card timing on the chat page.
- Message bottom actions and model-status text are clearer and less interruptive; membership upgrade entry display is refined.
Import, polish, stability
- Character card import is consolidated into one entry point.
- Unified interaction design across more surfaces.
- Framework-related crashes on specific phone models are addressed.
This is a substantial client update. Updating to the current store build is recommended. Later versions will keep tightening interaction details that earlier releases left rough.
Download
iOS and Android install links are on the site homepage: MiniTavern.
Tutorial: MiniTavern tutorial (English).
About the author
FAQ
Q1: Do I need both v1.5.0 and v1.5.1 separately?
No. Install or update to the latest store build. v1.5.1 sits on top of the v1.5.0 feature set and includes the critical fixes.
Q2: Can I still set a different model per character card?
Not through the old middle layer. Model selection is global for the app; use quick switch on the chat page when you need to change models mid-conversation.
Q3: Which local LLM setups are the OpenAI-format path aimed at?
Common OpenAI-compatible local servers such as KoboldCpp and LM Studio. Exact URLs and model IDs still depend on how you run those tools locally.
Written: August 3, 2026
Updated: August 3, 2026