← Models

Grok 4.3

xAI · Grok

xAI's recommended primary Grok caller and post-retirement redirect target for reasoning, non-reasoning, coding, and long-context agent work.

Part of Grok family · Other versions: Grok 4.20 , Grok 4.1 Fast , Grok Code Fast 1 , Grok 4
Type
multimodal
Context
1M tokens
Max Output
1M tokens
Status
current
Input
$1.25/1M tok
Output
$2.5/1M tok
API Access
Yes
License
proprietary
reasoning tool-use agentic multimodal long-context cost-efficient coding
Released April 2026 · Updated May 16, 2026

Overview

Freshness note: Model capabilities, limits, and pricing can change quickly. This profile is a point-in-time snapshot last verified on May 16, 2026.

Grok 4.3 is xAI’s current recommended primary API caller. It first appeared in the consumer Grok product in mid-April 2026 and rolled out across xAI’s API and partner integrations through the end of the month, with public pricing now standardized at a steep discount versus earlier flagship Grok tiers. xAI describes 4.3 as the most intelligent and fastest Grok model it has built, and the official guidance is to default to it for general API workloads outside specialized audio, image, and video endpoints.

That recommendation reframes the Grok lineup again after the May 2026 retirement notice. xAI now recommends Grok 4.3 as the migration target for Grok 4, Grok 4.1 Fast, Grok 3, and Grok Code Fast 1 callers, including coding and non-reasoning workloads that previously used separate model names. After May 15, deprecated text slugs redirect to Grok 4.3, with reasoning slugs routed at low effort and non-reasoning slugs routed at none.

Capabilities

Grok 4.3 is a reasoning model with extended chain-of-thought and a 1M-token context window. xAI emphasizes three capability areas in its current materials:

  • Agentic reasoning across long-context tool loops, including web search, X search, code execution, connectors, and MCP-style tool flows.
  • Multimodal input: text and image are first-class, and recent updates extended Grok’s input handling toward richer document and video understanding in product surfaces.
  • Document and presentation synthesis in product mode, with downloadable PDFs, spreadsheets, and slide decks coming out of conversation rather than as a separate generation step.
  • Coding and web-development workloads that previously routed to Grok Code Fast 1.

For developers, the practical change versus Grok 4 is not just price. It is that 4.3 is positioned as the API default, so routing logic that previously escalated to 4 for difficult work can simplify around 4.3 with selective fallbacks.

Technical Details

xAI’s developer documentation lists Grok 4.3 with a 1M-token context window, configurable reasoning, and support for non-reasoning mode. The migration guide documents four reasoning effort levels for this transition: none, low, medium, and high. The guide recommends Grok 4.3 directly for reasoning and coding workloads, and Grok 4.3 with reasoning_effort set to none for older non-reasoning Grok 4.1 Fast callers.

As with other xAI entries in this repo, maxOutput is stored as a documented context ceiling for comparison convenience. Applications should still set explicit per-request output limits and timeout budgets.

Pricing & Access

Listed API pricing (per 1M tokens):

  • Input: $1.25
  • Output: $2.50

This is a substantial reduction from the published Grok 4 rates of $3 input and $15 output per million tokens, and it pulls Grok 4.3 into a clearly cost-competitive lane against premium reasoning models from OpenAI and Anthropic. Cached input is discounted, and tool calls are billed separately.

Access options:

  • xAI API
  • Grok consumer apps and X ecosystem surfaces
  • xAI enterprise integrations
  • Partner inference and gateway providers where enabled

Best Use Cases

Use Grok 4.3 for:

  • General agentic and reasoning workloads where it is now xAI’s recommended default.
  • Long-context retrieval-heavy assistants that previously hit Grok 4 only for difficulty reasons.
  • Tool-using research workflows that benefit from xAI search, code execution, and MCP integrations.
  • Cost-sensitive production traffic that still needs strong reasoning behavior.
  • Coding and web-development agents migrating away from Grok Code Fast 1.

Use reasoning effort deliberately instead of routing new work through deprecated slugs. none is the migration path for old non-reasoning callers, low is the default redirect for old reasoning and coding slugs, and medium or high should be set explicitly when a workload benefits from deeper reasoning.

Comparisons

  • Grok 4 (xAI): Deprecated older reasoning route; 4.3 is the recommended replacement and post-retirement redirect target.
  • Grok Code Fast 1 (xAI): Deprecated coding route; 4.3 is the xAI migration target for coding and web-development workloads.
  • Claude Opus 4.7 (Anthropic): More expensive premium frontier route with stronger governance and platform breadth, while Grok 4.3 leans on tighter xAI ecosystem alignment and aggressive pricing.
  • GPT-5.5 (OpenAI): Comparable generalist flagship, with broader enterprise documentation and tooling ecosystem, while Grok 4.3 is the more cost-competitive tool-using reasoning option.