Release 0.22.0

Released 2017-01-28.

• Added per-model rate limits
• Fixed token counting for long prompts
• Reduced idle memory use

Release 853 · Release 854 · Release 91150

Back to the changelog