The models and pricing page lists per-token rates for Opus 4.6/4.7/4.8 but has no entry for any fast mode variant.
The model hosting page lists Claude Opus 4.6 (fast mode) (preview) as a supported Anthropic model, and it is selectable in the model picker. Under the previous billing system it was explicitly documented at a 30x request multiplier. Under the current token-based system its cost is undocumented.
Anthropic charges 2x for fast mode on their direct API ($10/$50 per MTok for Opus 4.8 vs $5/$25 standard). It is unclear whether Copilot passes this through, absorbs it, or has different terms.
Additionally, Opus 4.7 and 4.8 fast mode variants appear in the model picker via the API but are not listed on the model hosting page at all, unlike Opus 4.6 fast mode which is explicitly listed.
Please either:
- Add fast mode pricing entries to the pricing table, or
- Clarify that fast mode variants are billed at the same rate as their base models and document this explicitly
The models and pricing page lists per-token rates for Opus 4.6/4.7/4.8 but has no entry for any fast mode variant.
The model hosting page lists
Claude Opus 4.6 (fast mode) (preview)as a supported Anthropic model, and it is selectable in the model picker. Under the previous billing system it was explicitly documented at a 30x request multiplier. Under the current token-based system its cost is undocumented.Anthropic charges 2x for fast mode on their direct API ($10/$50 per MTok for Opus 4.8 vs $5/$25 standard). It is unclear whether Copilot passes this through, absorbs it, or has different terms.
Additionally, Opus 4.7 and 4.8 fast mode variants appear in the model picker via the API but are not listed on the model hosting page at all, unlike Opus 4.6 fast mode which is explicitly listed.
Please either: