[szhaomsft]

Summary

• Add GPT-5.6 Luna to text-model selections and the pipeline comparison, including BYOM cascade and GPT Live backend scenarios.
• Prefill verified South India Global short-context Standard inference rates: $0.20 input, $0.02 cached input, and $1.20 output per million tokens.
• Preserve exact meter names, IDs, effective dates, and a per-meter verification date of 2026-10-26 without changing the date of existing snapshot entries.
• Continue using Central India BYO speech references. Managed-mode Luna rows follow the existing explicitly labeled Global-model/BYO-speech economic reference, not a verified managed tariff.
• Document the pricing scope and add regression coverage for exact rates, provenance, cascade costs, and GPT Live backend costs.

Validation

• All 24 calculator tests pass; JavaScript syntax and whitespace checks pass.
• Browser verification confirms BYOM selection, editable rates, per-minute comparison, and GPT Live model preservation.
• Default Luna cascade: $0.022579/call ($0.015053/minute); GPT Live with Luna backend: $0.076519/call.

Deployment

The existing GitHub Pages workflow deploys the update automatically after merge into main. Unrelated local files are excluded.