v2026.7.10-2

Added

• Added the opt-in max thinking level across CLI, SDK, RPC, model selection, settings, and themes.
• Added request-wide input-token pricing tiers to custom model costs in models.json, modelOverrides, and extension-registered providers.
• Added ~ expansion for the shellPath setting (#6470 by @aaronkyriesenbach).

Changed

Fixed

• Fixed inherited post-compaction output-token budgeting to ignore stale assistant usage from before the compaction boundary.
• Fixed inherited GPT-5.4, GPT-5.5, and GPT-5.6 long-context pricing metadata, while excluding the nonexistent bare gpt-5.6 OpenAI/Azure alias.
• Fixed inherited Anthropic message conversion to preserve thinking blocks with empty thinking text but a valid signature instead of dropping them (#6457 by @davidbrai).
• Fixed inherited Anthropic replay to send the normalized thinking signature value after signature narrowing.
• Fixed inherited OpenAI-compatible simple streams to preserve provider-specific max reasoning mappings instead of clamping them.
• Fixed inherited same-model signed thinking replay to keep non-empty signed blocks only when provider state must be replayed, while preserving empty opaque signed blocks.

Removed