Release 0.22.6

Released 2017-01-04.

• Added per-model rate limits
• Improved streaming reliability
• Fixed token counting for long prompts

Release 81859 · Release 81860 · Release 92793

Back to the changelog