xAI retired eight Grok API model slugs on May 15, 2026, at noon Pacific Time. The company's migration guide lists grok-4-1-fast-reasoning, grok-4-1-fast-non-reasoning, grok-4-fast-reasoning, grok-4-fast-non-reasoning, grok-4-0709, grok-code-fast-1, grok-3, and grok-imagine-image-pro.
Official docs say those slugs continue to resolve after the cutoff. Reasoning IDs in the list are served by grok-4.3 with low reasoning effort. Non-reasoning IDs go to grok-4.3 with none. grok-code-fast-1 routes to grok-build-0.1. grok-imagine-image-pro routes to grok-imagine-image-quality.
xAI says requests that stay on a deprecated text slug after May 15 are billed at grok-4.3 rates of $1.25 per 1 million input tokens and $2.50 per 1 million output tokens, not at the retired model's price. The live models page shows the same $1.25 / $2.50 pair for grok-4.3 when the prompt is under 200,000 tokens, and $2.50 / $5.00 when the prompt is longer.
The company recommends changing the "model" field before May 15 rather than riding the redirect, so teams control reasoning effort and cost. For reasoning workloads it points developers to grok-4.3 with an explicit effort setting. For coding it points to grok-build-0.1. Company materials also say grok-4.3 has a 1 million-token context window and four effort levels: none, low, medium, and high.
Independent developer write-ups match the silent-redirect reading. A DEV Community post last verified Aug. 7, 2026, against xAI's docs says the eight retired slugs still work and bill at grok-4.3 rates. Second Talent's mid-2026 Grok lineup note likewise says May retirements now redirect to Grok 4.3.
Some secondary coverage and customer-email accounts framed the cutoff as a hard stop where old IDs would no longer work. That reading conflicts with the migration guide's redirect language. Until a vendor changes the docs, treat silent redirect plus grok-4.3 billing as the documented behavior, and pin an explicit model and reasoning effort if bill or quality control matters.