v1.1.0

fast is now the default jury preset.

The default jury changes from three models to two:

	Before (balanced)	After (fast)
Jury	GPT-5.6 Terra max · Grok 4.5 high · Kimi K3 max	GPT-5.6 Luna low · DeepSeek V4 Flash low
Referee / verifier	kimi-k3	deepseek-v4-flash-0731

Lower latency and cost by default. If you relied on the previous default, pin it explicitly — nothing else changes:

# .juror.yml
preset: balanced

or --preset balanced on the CLI, or preset: balanced on the Action.

Minor rather than patch because this changes behaviour for anyone who had not pinned a preset. This release also reconciles the npm package version with the GitHub release version.