AI answers no longer get cut off, and the model list is live
Longer replies from the AI assistant used to stop mid-sentence at a fixed token limit, and the model menu was a short, hardcoded list. 0.8.6 removes the cap and lets each provider tell texspark what it can do.
- Answers run to completion. The shared response cap is gone. OpenAI and Gemini now use their full output budget, so long explanations and code blocks finish instead of getting truncated.
- Each Claude model gets its real limit. Anthropic’s API requires a max, so texspark reads each model’s true output ceiling from the provider and sends exactly that — never over the limit, never needlessly short.
- The model picker shows what your key can actually use. Preferences → AI and the assistant header now list models fetched live from each provider, so new releases appear on their own. Offline, it falls back to a built-in list.
Leave a Reply