[Cahl-Dee]

Description

Follow-up to #1291. The Grid guide listed seven example instruments. GET https://api.thegrid.ai/v1/models returns 17, so this lists all of them.

• Replace the example table in fern/customization/custom-llm/thegrid.mdx with two tables: the nine capability tiers (text, agent and code, each in standard, prime and max) and the eight *-latest aliases that follow the newest release from one lab.
• Use the one-line description we maintain for each instrument.
• Keep the link to GET /v1/models, the instruments page and pricing. Prices and token limits are not in the table because they change.
• Add a short "Check the other model settings" table for the Custom LLM dashboard fields (Model, URL, Max Tokens, Request Timeout, Metadata Send Mode). It points to GET /v1/models for the per-instrument limits instead of copying numbers.
• Change the example assistant from text-standard to text-prime
• No navigation, provider overview or other page changes.

Testing Steps

• Run node scripts/fern/run.cjs check (Fern 5.151.1): 0 errors, and none of the existing warnings mention this page
• Open the preview deployment and check that the two instrument tables and the settings table render on The Grid page
• Check the instrument IDs against GET https://api.thegrid.ai/v1/models

We also ran the guide's setup end to end through Vapi's custom-llm provider on 2026-10-21: a credential, an assistant and a chat reply for text-prime, agent-prime, code-prime and claude-opus-latest, all passing.