Daily vendor refresh (2026-06-09): OpenAI nano + Mistral Large 3 pricing

[sciencebased]

Summary\n- Update Mistral Large 3 pricing to /bin/bash.50 input / .50 output per 1M tokens, with batch input at /bin/bash.25 (50% discount). Source: https://mistral.ai/pricing\n- Add pricing row for gpt-5.4-nano (/bin/bash.20 input / /bin/bash.02 cached input / .25 output per 1M tokens). Source: https://platform.openai.com/docs/pricing\n- Add supporting research report: research/research_report_2026-06-09.md\n\n## Notes\n- Did not update Gemini pricing because the official pricing page currently shows many models as "Free of charge" without a stable paid-tier table; treated as disputed/unclear (see research report).\n\n## Checks\n- [x] npm run typecheck\n- [x] npm run build\n

[sciencebased]

Escalation 4753e843 — re-verification (2026-06-11)

Handled the escalated daily run for 2026-06-11. No new PRs opened. Re-verified PR #10's existing data edits against primary vendor pages today; both are correct, so no src/data changes were made this run. Added a dated research addendum (research/research_report_2026-06-11.md, commit de54b5b).

Verified against primary sources (still current today):

• OpenAI gpt-5-4-nano → $0.20 input / $0.02 cached / $1.25 output per 1M tokens. Matches the row already on this branch. Source: https://platform.openai.com/docs/pricing (redirects to https://developers.openai.com/api/docs/pricing). The dataset's dash-ID gpt-5-4-nano (display "GPT-5.4 nano", models.ts) maps to OpenAI's gpt-5.4-nano. No model was renamed and no ID invented — the ID already exists in the catalog.
• Mistral mistral-large-3 → $0.50 input / $1.50 output, batch $0.25 (50% discount). Matches the row already on this branch. "Mistral Large 3" maps to existing mistral-large-3. Source: https://mistral.ai/pricing.

Held back (mapping ambiguity / out of scope):

• OpenAI gpt-5.5-pro / gpt-5.4-pro ($30/$180 per 1M): no corresponding entry in src/data/models.ts, and the pricing page alone doesn't establish stable model metadata. Adding price rows would create dangling modelIds. Deferred to a dedicated catalog-expansion change.
• xAI Grok Build 0.1: lacks required canonical catalog metadata; documented only.
• Google Gemini: paid-tier table still renders inconsistently ("Free of charge"); held until stable.

No overlap with PR #9 (Cohere Command R+ 08-2024) — this branch does not touch Cohere; command-r-plus-2025 remains $3/$15.

Checks run (clean working tree, build artifacts reverted):

• npm run typecheck (tsc -b --noEmit) — passed.
• npm run build (tsc -b && vite build) — succeeded (178 modules).

Files changed this run: research/research_report_2026-06-11.md (added). Commit: de54b5b.