#17602: fix: correct MiniMax M2.5 pricing constants (75x overcharge)

## Summary - Problem: MiniMax M2.5 pricing constants were ~75x higher than actual official pricing, causing `/usage cost` to report wildly inflated costs (e.g., $2.25 instead of $0.03) - Why it matters: Users on MiniMax models see incorrect cost tracking, eroding trust in usage reporting - What changed: Updated `MINIMAX_API_COST` to match [official MiniMax pricing](https://platform.minimaxi.com/document/Price), added separate `MINIMAX_LIGHTNING_API_COST` for Lightning models, extended `buildMinimaxModel()` to accept optional `cost` parameter - What did NOT change (scope boundary): No changes to cost calculation logic, display formatting, or other provider pricing ## Change Type (select all) - [x] Bug fix - [ ] Feature - [ ] Refactor - [ ] Docs - [ ] Security hardening - [ ] Chore/infra ## Scope (select all touched areas) - [ ] Gateway / orchestration - [ ] Skills / tool execution - [ ] Auth / tokens - [ ] Memory / storage - [ ] Integrations - [ ] API / contracts - [ ] UI / DX - [ ] CI/CD / infra > Note: Affects `agents` (model config) and `commands` (onboarding auth models) — pricing constants only. ## Linked Issue/PR - Closes #17548 ## User-visible / Behavior Changes - `/usage cost` now reports accurate costs for MiniMax models - Example: A session previously reported as ~$2.25 will now correctly show ~$0.03 - Lightning models (`MiniMax-M2.1-Lightning`, `MiniMax-M2.5-Lightning`) now have separate accurate pricing ($2.40/1M output vs $1.20/1M standard) ## Security Impact (required) - New permissions/capabilities? `No` - Secrets/tokens handling changed? `No` - New/changed network calls? `No` - Command/tool execution surface changed? `No` - Data access scope changed? `No` ## Repro + Verification ### Environment - OS: Ubuntu 24.04 - Runtime/container: Node.js v22.22.0 - Model/provider: MiniMax M2.5 / M2.1-Lightning / M2.5-Lightning - Integration/channel: N/A - Relevant config: Any config using MiniMax models ### Steps 1. Configure a MiniMax model as primary or fallback 2. Run a session and check `/usage cost` 3. Before fix: costs reported ~75x higher than actual 4. After fix: costs match [official MiniMax pricing](https://platform.minimaxi.com/document/Price) ### Expected - Cost per 1M tokens: Input $0.30, Output $1.20 (standard) / $2.40 (Lightning) ### Actual (before fix) - Cost per 1M tokens: Input $15, Output $60 ## Evidence - [x] TypeScript compilation: 0 errors - [x] Linting (oxlint): 0 warnings, 0 errors - [x] Formatting (oxfmt): Applied - [x] Pricing verified against [official MiniMax documentation](https://platform.minimaxi.com/document/Price) ## Human Verification (required) - Verified scenarios: TSC compilation, lint, format pass; pricing constants cross-referenced with official MiniMax pricing page - Edge cases checked: Lightning vs standard model separation, cache read/write pricing, both duplicate constant locations (`models-config.providers.ts` and `onboard-auth.models.ts`) - What you did **not** verify: Live MiniMax API session cost comparison (verified via pricing documentation) ## Compatibility / Migration - Backward compatible? `Yes` - Config/env changes? `No` - Migration needed? `No` ## Failure Recovery (if this breaks) - How to disable/revert this change quickly: Revert the two files - Files/config to restore: `src/agents/models-config.providers.ts`, `src/commands/onboard-auth.models.ts` - Known bad symptoms: Incorrect cost reporting for MiniMax models ## Risks and Mitigations - Risk: Pricing may change if MiniMax updates their rates - Mitigation: Values sourced directly from official documentation; same risk exists for all hardcoded provider pricing --- 🤖 AI-assisted (Opus 4.6 implementation, reviewed by Opus 4.6 + GPT 5.3 Codex with thinking high). Fully tested (TSC + lint). I understand what the code does. <!-- greptile_comment --> <h3>Greptile Summary</h3> Fixes MiniMax model pricing constants that were ~75x higher than actual official rates, causing `/usage cost` to report wildly inflated costs. Both duplicate constant locations (`models-config.providers.ts` and `onboard-auth.models.ts`) are updated to match [official MiniMax pricing](https://platform.minimaxi.com/document/Price). - Updated `MINIMAX_API_COST` from $15/$60 (input/output per 1M tokens) to the correct $0.30/$1.20 - Added separate `MINIMAX_LIGHTNING_API_COST` ($0.30 input / $2.40 output) for Lightning models (`MiniMax-M2.1-Lightning`, `MiniMax-M2.5-Lightning`) - Extended `buildMinimaxModel()` and `buildMinimaxTextModel()` with optional `cost` parameter, defaulting to standard pricing - Cache pricing also corrected: `cacheRead` from $2 to $0.03, `cacheWrite` from $10 to $0.375 - No changes to cost calculation logic, display formatting, or other provider pricing <h3>Confidence Score: 5/5</h3> - This PR is safe to merge — it only changes hardcoded pricing constants with no logic or control flow modifications. - The changes are limited to correcting numerical pricing constants in two files and adding a new cost constant for Lightning models. The optional `cost` parameter addition to builder functions is backward-compatible (defaults to existing behavior). Both constant locations are updated consistently. The onboarding path (`buildMinimaxApiModelDefinition`) doesn't currently handle Lightning models, which is a minor gap but not a regression since those model IDs are not passed through onboarding flows. - No files require special attention _{Last reviewed commit: 85e6892} <!-- greptile_other_comments_section --> _{(2/5) Greptile learns from your feedback when you react with thumbs up/down!} <!-- /greptile_comment -->