Grok 4.3

Grok 4.3 is an xAI model positioned as a unifying successor within the Grok family. In xAI's May 15, 2026 developer migration documentation, Grok 4.3 is explicitly named as the redirect target for several prior Grok endpoints that were retired on that date, including the Grok 4 fast reasoning and non-reasoning variants, grok-4-0709, grok-3, and related slugs. Reasoning-mode requests land on Grok 4.3 with low reasoning effort, while non-reasoning requests are served with reasoning disabled, making it the practical fallback for nearly the entire previous Grok lineup on the xAI API. This consolidating role signals a deliberate shift by xAI toward a single flagship reasoning model rather than a sprawling set of variants.

The model's fit is shaped by its benchmark placement and a sharp pricing reset. Reporting from Fello AI records Grok 4.3 scoring 53 on the Artificial Analysis Intelligence Index, placing it above Claude Sonnet 4.6 but below GPT-5.5, Claude Opus 4.7, and Gemini 3.1 Pro Preview, which frames it as a mid-frontier rather than top-tier option. The release followed an earlier $300-per-month SuperGrok Heavy beta, with broader API availability around April 30, 2026, and an input price roughly 40% lower than Grok 4.20 alongside an output price cut closer to 60%. Together, those moves position Grok 4.3 as a cost-conscious workhorse for developers who want Grok-style reasoning and tool use without paying premium-frontier rates.

xAIgrok-4.3grok

Open provider page Provider docs

Quick Info

Provider: xAI
Model key: grok-4.3
Release date: Apr 17, 2026
Last updated: Apr 17, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $1.25
Output token cost: $2.50

Limits

Output tokens: 30,000 tokens
Context window: 1,000,000 tokens

Transparent token rates

Compare Grok 4.3 pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Grok 4.3

xAI

CoverageRelease Notes

Complete timeline of Grok updates from xAI, including the SpaceX acquisition, Grok 4.3, Imagine video generation, and enterprise API launches.

302.AI

CoverageBenchmark

Coursiv's coding-focused review highlights Grok 4.3 as a long-context, low-cost option for codebase reading, debugging, test generation, and API automation, citing a 1 million-token context window and xAI API support. It frames Grok 4.3 alongside the Grok Build terminal coding agent and real-time web/X research integra The same article recommends caution versus Claude Code or Codex for high-stakes multi-file refactors, mature coding-agent workflows, and ecosystem depth, and positions Grok 4.3 as useful as a second-pair-of-eyes code reviewer. A July 24, 2026 update note states that Grok 4.5 is now xAI's flagship coding model, but expl

xAI

CoverageBenchmark

xAI launched Grok 4.3 on April 30, 2026, with notable gains in agentic task performance and sharply reduced API pricing relative to the prior Grok 4.20 0309 v2 release, according to Artificial Analysis's launch analysis. The model reaches a score of 53 on the Artificial Analysis Intelligence Index, placing it just abov The most significant benchmark jump is on the real-world GDPval-AA evaluation, where Grok 4.3 posts an ELO of 1500, a 321-point gain over Grok 4.20 0309 v2, surpassing Gemini 3.1 Pro Preview, Muse Spark, GPT-5.4 mini (xhigh), and Kimi K2.5 while still trailing GPT-5.5 (xhigh) by 276 ELO. Pricing is reduced 37.5% on inp

xAI

CoverageBenchmark

BenchLM's profile for Grok 4.3, last verified on July 26, 2026, confirms the April 30, 2026 release date and now flags the model as superseded by xAI's newer Grok 4.5 line. The aggregator reports an overall score of 64.2/100, ranking Grok 4.3 35th of 215 on its public leaderboard and 29th of 101 on the verified leaderb For developers, the profile lists API pricing at $1.25 per million input tokens and $2.50 per million output tokens, a 1M-token context window, and throughput of 209 tokens per second for this proprietary reasoning model. BenchLM's own evidence coverage is thin, with only 25 of 369 tracked benchmarks reported and just