Grok 4.3 Launches, Claude Security Beta Opens, Cursor-xAI Deal Emerges

Three significant AI developments landed within days of each other in late April 2026, reshaping the competitive landscape for developers, enterprises, and investors. xAI launched Grok 4.3 with measurable cost reductions, Anthropic opened Claude Security to public beta for enterprise customers, and SpaceX announced a partnership with Cursor that includes an option to acquire the coding platform for $60 billion.

The Grok 4.3 release represents xAI's most aggressive pricing move to date. According to benchmark analysis from Artificial Analysis, the model achieves a score of 53 on the Intelligence Index while costing approximately 20% less to run the full benchmark suite than Grok 4.20. Input token prices dropped 37.5% and output token prices fell 58.3%, making it one of the lowest-cost models at its intelligence level.

The most dramatic improvement appears in agentic task performance. Grok 4.3 scored an ELO of 1500 on GDPval-AA, up 321 points from Grok 4.20's 1179. That places it ahead of Gemini 3.1 Pro Preview, Muse Spark, and GPT-5.4 mini (xhigh), though it still trails GPT-5.5 (xhigh) by 276 Elo points. The model also maintains strong instruction-following capabilities with a 98% score on ℑ²-Bench Telecom and 81% on IFBench.

xAI's official release notes detail additional features beyond raw benchmarks. Grok 4.3 now supports video uploads (mp4, webm, mov formats with inline preview), improved spreadsheet and document creation, and an agent library for specialized tasks. The knowledge cutoff is December 2025, and the model is rolling out to SuperGrok and Premium+ subscribers. The company states it's working toward giving Grok full access to internet, files, email, and calendar services — essentially treating the model like a human sitting down to do real work.

There's a tradeoff worth noting: Grok 4.3 uses approximately 44% more output tokens than Grok 4.20 to run the Intelligence Index, and its AA-Omniscience Non-Hallucination Rate dropped 8 points compared to the previous version. Whether the cost savings justify the slightly higher hallucination rate depends entirely on your use case (developers chasing benchmarks will care; enterprise users might not).

Anthropic's Claude Security launch takes a different approach, targeting enterprise cybersecurity workflows rather than general intelligence. The tool, now in public beta for Claude Enterprise customers, leverages the Opus 4.7 model to scan codebases for vulnerabilities and generate targeted patches. It requires no API integration or custom agent build — users can access it directly from the Claude.ai sidebar or at claude.ai/security.

The product addresses three specific pain points security teams reported during testing: false positives, time from scan to fix, and the need for ongoing coverage rather than one-off audits. Anthropic added a multi-stage validation pipeline that independently examines each finding before it reaches an analyst, attaching a confidence rating to every result. Teams can schedule scans to maintain regular cadence, target specific directories within repositories, and export findings as CSV or Markdown for existing tracking systems.

SecurityWeek reported that Claude Security has already been tested by hundreds of organizations of all sizes in limited research preview. The tool traces data flows across files and modules, reads source code, and explains how components interact — reasoning about code much like a security researcher rather than searching for known patterns. It generates instructions for targeted patches that can be worked through with Claude Code on the Web.

Anthropic is also integrating Opus 4.7's capabilities into platforms from CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, TrendAI, and Wiz. Consulting firms including Accenture, BCG, Deloitte, Infosys, and PwC are deploying Claude-integrated solutions for vulnerability management and incident response. Access for Claude Team and Max customers is coming soon, according to the company.

The Cursor-xAI deal reveals the business strategy behind these technical launches. TechCrunch reported that SpaceX struck a deal with Cursor to develop a next-generation coding and knowledge work AI, with an option to buy the startup for $60 billion later this year. The partnership combines Cursor's product and distribution to expert software engineers with SpaceX's Colossus supercomputer, which the company claims has the equivalent compute power of a million Nvidia H100 chips.

Either way — whether SpaceX pays Cursor $10 billion for its work or acquires the company for $60 billion — the move addresses weaknesses at both companies. Neither Cursor nor xAI has proprietary models that can match leading offerings from Anthropic and OpenAI. Cursor still uses and sells access to Claude and GPT models even as both firms roll out their own coding tools, an awkward arrangement this partnership may be designed to eventually escape.

The valuation trajectory is staggering. Cursor was valued at $2.5 billion in January 2025, climbed to $9 billion by May 2025, and reached a $29.3 billion post-money valuation when it closed $2.3 billion in Series D funding in November. Last week, TechCrunch reported Cursor was eyeing a $50 billion valuation in an upcoming private fundraising round. The $60 billion option price reflects an astonishing series of leaps.

Two of Cursor's most senior engineering leaders, Andrew Milich and Jason Ginsberg, left the company last month to join xAI, where both report directly to Musk. xAI will begin renting computing power from its data centers to Cursor, with the coding startup using tens of thousands of xAI chips to train its latest AI model. The deal won't shock those who follow the industry closely, but it does reveal the extent of Musk's increasingly sprawling tech conglomerate.

Whether users actually pay for these improvements remains the real question. Grok 4.3's cost efficiency matters if developers choose it over competitors. Claude Security's value depends on whether enterprises trust AI-generated vulnerability findings enough to act on them. The Cursor-xAI deal hinges on whether SpaceX can successfully integrate a consumer-facing developer tool into its infrastructure-heavy business model. None of these developments guarantee success — they're just expensive bets on where the market might go.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn