xAI Launches Grok 4.3 Beta With 1 Trillion Parameters

xAI Ships Grok 4.3 Beta With 1 Trillion Parameters

Elon Musk's xAI released Grok 4.3 beta on April 17, 2026, making it available across iOS, Android, and web for SuperGrok Heavy subscribers paying $300 per month. The model doubles the parameter count of its predecessor Grok 4.20, reaching 1 trillion parameters, and introduces a suite of new capabilities including native video input, sharper long-context reasoning, and enhanced multimodal handling.

The release comes as competition in the frontier model space has intensified sharply, with Anthropic shipping Claude Opus 4.7 just one day earlier and OpenAI continuing to iterate on GPT-5.4. Musk announced that xAI intends to ship improvements to Grok 4.3 "almost every single day," signaling an unusually aggressive cadence for a frontier model that most labs update on weekly or monthly cycles.

Native Document Generation and Video Understanding

Grok 4.3 introduces capabilities that no previous Grok version offered: native PDF creation, PowerPoint slide generation, spreadsheet output, and video input processing. These features position Grok as a more complete productivity tool rather than a pure conversational AI, directly competing with the document-generation capabilities that Microsoft Copilot and Google Gemini have built into their enterprise suites.

The video input capability allows users to feed video clips directly into Grok for analysis, summarization, and question-answering — a feature that Google Gemini pioneered but that most competing models have been slow to match at production quality. Combined with the model's enhanced long-context handling, Grok 4.3 can process substantially more information in a single session than its predecessor.

The $300 Paywall Draws Scrutiny

Early access to Grok 4.3 is locked behind xAI's SuperGrok Heavy tier at $300 per month — ten times the cost of the standard $30 SuperGrok subscription. Regular SuperGrok subscribers can see the model listed in their model selector with an "Early Access" label but cannot activate it. xAI has not provided a timeline for when Grok 4.3 will become available at the standard tier.

"Grok 4.3 is now available as an early beta. We intend to ship improvements to it almost every single day." — Elon Musk

The pricing strategy mirrors a broader industry trend toward tiered access for frontier models. Anthropic restricts its most capable Mythos model to vetted enterprise partners, while OpenAI charges premium rates for its highest-capability API tiers. But xAI's approach of gating access purely behind a consumer subscription price — rather than enterprise agreements or safety reviews — has drawn mixed reactions from the developer community.

Grok Build and CLI: Entering the Coding Agent Race

Alongside the Grok 4.3 release, Musk publicly set a next-week timeline for the debut of Grok Build and Grok CLI, marking xAI's formal entry into the agentic coding category. Grok Build is shaping up as a dual-track offering that lets users run coding tasks either locally via a CLI-backed agent or remotely through a web interface, with support for up to eight concurrent AI coding agents running simultaneously on a single project.

The move pits xAI directly against Claude Code, OpenAI's Codex, Google's Jules, Cursor, and GitHub Copilot — tools that have dominated developer mindshare throughout the first half of 2026. For engineers evaluating their toolchain options, the question will be whether Grok 4.3's raw parameter count and xAI's rapid iteration cadence can translate into practical coding assistance that matches or exceeds the more established alternatives.

What This Means for Engineers and Job Seekers

The rapid proliferation of frontier coding models — now spanning offerings from xAI, Anthropic, OpenAI, Google, and multiple IDE-native tools — is reshaping what employers expect from software engineering candidates. Familiarity with AI-assisted development workflows is increasingly a baseline expectation rather than a differentiator. For engineers currently job-seeking, hands-on experience with multiple AI coding assistants, and the judgment to know when to use each, is becoming as important as traditional algorithmic skills.

The $300 monthly price tag for Grok Heavy may limit individual developer adoption, but enterprise teams evaluating AI tooling budgets are increasingly treating these costs as equivalent to a software license — a small fraction of an engineer's compensation that can yield outsized productivity gains if the tool delivers on its promises.