Problem#
These chat models have pricing: null in the native catalog (?pricing=1 and GET /v1/models/{provider}/{id}):
| Provider | Chat rows without a card | Examples |
|---|
| openai | 6 / 106 | gpt-6-luna, gpt-6-sol, gpt-5-search-api |
| gemini | 18 / 43 | antigravity-preview-09-2026, deep-research-preview-04-2026, gemini-2.5-flash-native-audio-latest |
| byteplus | 13 / 13 | deepseek-v4-1-flash-260910, glm-5-3-flash-260828, dola-seed-2-1-turbo-260628 |
| groq | all 11 rows | allam-2-7b |
| mistral | all 52 rows | codestral-2508 |
gpt-6-luna and gpt-6-sol are the newest OpenAI models. gpt-6-astra in the same family has a docs-derived card from developers.openai.com/api/docs/models/gpt-6-astra.md. OpenRouter has prices for openai/gpt-6-luna ($0.10 / $0.50) and openai/gpt-6-sol ($2 / $10).
Impact#
TanStack AI syncs model-meta.ts from the native catalog. When there is no card, we fall back to the OpenRouter row. BytePlus, Groq, and Mistral have no OpenRouter fallback, so those models ship with no price.
Ask#
- Fill cards for
gpt-6-luna and gpt-6-sol from their docs pages, the same way as for gpt-6-astra.
- Add a pricing source for BytePlus, Groq, and Mistral.
- Check why the 18 Gemini chat rows have no card.
Acceptance#
GET /v1/models/openai/gpt-6-luna returns a rate card with factSources.pricing.
GET /v1/status shows BytePlus, Groq, and Mistral with a non-zero counts.priced.