LinguaAI Studio

The newest Claude models, batch stories that cache, and Transcribe that works on them. - Claude Haiku 5.5, Sonnet 5.5 and Opus 5.5 in chat, Studio, Shortcuts and on the Watch. Sonnet 5.5 and Haiku 5.5 think by default and take every effort level; Off means the least thinking the model allows, on stories, transforms, batches and the chat. Claude Sonnet 4.5 is gone; anything pinned to it runs on Sonnet 5.5. - Studio no longer gets closed by iOS while it sits in the background: the chat database moves to Studio's own storage once, on first launch, after a check that everything arrived. Nothing changes in iCloud. - Studio's batch mode caches the stable part of the prompt and reads it for the rest of the batch at a tenth of the price. The notification says what the cache did. A batch whose first story landed while the rest kept spinning now sends the rest straight after. - Transcribe: cleanup by Claude works on the newest models again, and ElevenLabs Scribe says why when it refuses instead of sitting on Listening. - Perplexity runs as a router now: one Perplexity key gives you Claude, GPT, Gemini and GLM at the makers' prices, plus Sonar for web search. - Mistral Large 4 (preview) is in the pickers; Magistral Medium is gone, Mistral answers it with Mistral Medium. - A story Anthropic's safety filter stops says so, and which filter. The "previously written" list carries each earlier title in full. - No prompt can go invisible: a prompt whose category was deleted or has not arrived yet sits under No category. The backup shortcut includes the Studio presets with their steering text.