Grok 4 Fast

Retired · Language, Reasoning

Grok 4 Fast is an xAI reasoning model that xAI describes as cost-efficient, released on 19 September 2025 in the Grok apps, including for free users, and on the xAI API with a 2-million-token context window. One set of weights serves both reasoning and non-reasoning responses, offered on the API as two model ids; both were retired on 15 May 2026. [1] [2] [3] Primary source

Timeline of Grok 4 Fast →

Claims and evidence

  • Released Primary source[1]header; availability sections
  • Retired Effective 12:00 PM PT on the xAI API. Primary source[3]first paragraph
  • Status Retiredgrok-4-fast-reasoning and grok-4-fast-non-reasoning retired from the xAI API; requests are served by grok-4.3. Primary source[3]list of retired models
  • Replaced by Grok 4.3The reasoning slug redirects to grok-4.3 with low reasoning effort, the non-reasoning slug with reasoning effort none. Primary source[3]Recommended Replacements
  • Successor of Grok 3 mini EditorialEditorial link along the cost-efficient tier: the announcement compares Grok 4 Fast with Grok 3 Mini and says it builds on learnings from Grok 4, without naming a predecessor or a derivation method. Primary source[1]Advancing Cost-Efficient Intelligence
  • Change · Architecture Reasoning and non-reasoning modes run on the same model weights, steered by the system prompt, instead of separate models.Compared with Grok 3 mini Primary source[1]Unified Model: Reasoning and Non-Reasoning
  • Change · Tool use Trained end to end with tool-use reinforcement learning to decide when to call tools such as code execution and web or X search.Compared with Grok 3 mini Primary source[1]Native Tool Use with SOTA Search
  • Change · Availability Offered to all Grok users including free users; xAI says this gave free users its latest model without restrictions for the first time.Compared with Grok 3 mini Primary source[1]Grok 4 Fast in grok.com, iOS, and Android apps
  • Change · Efficiency xAI reports results comparable to Grok 4 with substantially fewer thinking tokens on average.Compared with Grok 4 Primary source[1]
  • Feature Reasoning modeOne set of weights handles reasoning and non-reasoning modes, steered by the system prompt; offered on the API as two model ids. Primary source[1]Unified Model: Reasoning and Non-Reasoning [2]1 Introduction
  • Feature Tool useTrained end to end with tool-use reinforcement learning, including web and X search. Primary source[1]Native Tool Use with SOTA Search
  • Context window 2M tokens Primary source[1]first paragraph; xAI API section
  • Parameters Not disclosedNot disclosed by xAI. No claim madeNo source
  • Access consumer app, API Primary source[1]Grok 4 Fast in grok.com, iOS, and Android apps; Grok 4 Fast on OpenRouter, Vercel AI Gateway, and the xAI API

Lineage

Predecessors

Successors

Based on

Not derived from another model.

Variants and derived

None recorded.

Siblings

None recorded.

All ancestors

All descendants

Variants

No variants recorded in this record.

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from xAI on AI Radar →