xAI Grok 4.3 Goes Public: $1.25 Per Million Tokens and Always-On Reasoning Shake Up the LLM Market

Grok 4.3 Drops for Everyone — No $300 Subscription Required

On April 30, 2026, xAI flipped the API switch on Grok 4.3's full public release with almost no fanfare. There was no press conference, no launch event — just a small banner in the developer console and a pricing page update. After spending two weeks as an exclusive perk for $300-per-month SuperGrok Heavy subscribers, the model is now available to any developer at a price point that is turning heads across the industry.

At $1.25 per million input tokens and $2.50 per million output tokens, Grok 4.3 is priced at roughly one-quarter the cost of Anthropic's Claude Opus 4.7 ($5/$25) and less than half the cost of OpenAI's GPT-5.5 ($5/$30). According to analysis from Artificial Analysis, the pricing reflects a 40% reduction in input costs and a 60% reduction in output costs compared to Grok 4.20, xAI's previous flagship.

Key Features

1. Always-On Reasoning

Unlike models that allow developers to toggle reasoning mode on or off, Grok 4.3 runs with reasoning permanently active. xAI's documentation describes this as a deliberate architectural choice: the model is "intended for agentic, instruction-following, and long-document analysis workloads" where the overhead of reasoning steps is consistently justified. For users who want a lightweight, fast model without reasoning overhead, Grok 4.3 is not the right tool — but for complex task execution it means consistent depth without configuration.

2. One Million Token Context Window

Grok 4.3 ships with a 1 million-token context window, matching the upper tier of current frontier models. This is suitable for processing entire codebases, lengthy legal documents, multi-session research threads, or large dataset analyses without chunking. The pricing above 200,000 tokens is billed at higher per-token rates, so very-long-context workloads will cost more than the headline price implies.

3. Custom Voices Suite

Alongside the model release, xAI launched Custom Voices, a voice cloning suite integrated into the Grok platform. Users can create up to 30 custom voices simultaneously. Each custom voice is private to the creating team and is never made available to other users, providing a commercial license assurance for enterprise use. The base Text-to-Speech service offers five preset voices at $4.20 per million characters, and the custom voice feature allows teams to clone their own voice signatures for audio products, customer service automation, and AI-narrated content.

4. Agentic Output Generation

Grok 4.3 includes a built-in environment that can write and execute code and produce downloadable files. Users can prompt the model to generate fully populated spreadsheets, PowerPoint-compatible decks, and PDF documents directly from conversation. This extends Grok 4.3's utility beyond text generation into document production workflows, reducing the need for post-processing steps.

5. Competitive API Pricing Strategy

xAI's pricing has triggered immediate discussion in developer communities. As one X post from BridgeMind AI put it: "Anthropic: $5/$25. OpenAI: $5/$30. xAI: $1.25/$2.50. The pricing wars are getting brutal." The price differential is real, though benchmark comparisons matter. Independent testing from Robo Rhythms found that Grok 4.3 does not consistently outperform Claude Opus 4.7 across all task types, suggesting the value proposition is one of cost efficiency rather than raw capability leadership.

Usability Analysis

For developers already familiar with the xAI API, Grok 4.3's full release removes the previous barrier of the $300/month SuperGrok Heavy subscription. The model is accessible via the grok-4-3-20260430 model identifier and integrates with standard OpenAI-compatible API patterns, lowering the friction for teams switching from or testing alongside other providers.

The always-on reasoning mode means latency will be higher than a non-reasoning model. Developers who need sub-second responses for simple classification or extraction tasks should benchmark carefully before committing. For complex, multi-step agentic tasks — code generation, document analysis, structured research — the always-on reasoning is well-suited.

The Custom Voices suite is a meaningful addition for startups building voice-first products who want branded voice experiences without building a separate TTS pipeline.

Pros and Cons

Pros:

• Industry-disrupting price point: $1.25/$2.50 per million tokens
• 1M token context window for large document workloads
• Always-on reasoning provides consistent depth for complex tasks
• Custom Voices suite with private team licensing
• Native document/spreadsheet/PDF generation
• Full API access without premium subscription requirement

Cons:

• Always-on reasoning increases latency — not suitable for speed-critical applications
• Higher per-token rates above 200K context length
• Benchmark performance does not consistently exceed Claude Opus 4.7
• No opt-out for reasoning mode limits flexibility
• Memory functionality remains absent from Grok 4.3

Outlook

xAI's aggressive pricing with Grok 4.3 is likely to accelerate the broader pricing compression already underway in the frontier model market. When a model with 1M context and always-on reasoning launches at $1.25 per million input tokens, it forces every other lab to justify their premium pricing through demonstrably superior capability or ecosystem value.

The Custom Voices launch signals that xAI is building a broader platform, not just a model API. Voice generation, agentic document output, and code execution are converging in Grok 4.3 into a platform that competes with OpenAI's growing suite of capabilities. If xAI can demonstrate that Grok 4.3's reasoning quality is genuinely competitive with Opus 4.7 and GPT-5.5 across enterprise workloads, the price-performance gap it offers becomes a serious consideration for cost-conscious deployment teams.

xAI's rapid release cadence — beta to full public release in two weeks — also suggests that Grok 4.4 or improvements to Grok 4.3 are not far off.

Conclusion

Grok 4.3's full public release is primarily a story about economics. At $1.25 per million input tokens with a 1M context window and always-on reasoning, xAI is making a credible case for consideration in any cost-sensitive enterprise deployment. Teams building agentic workflows, long-document analysis tools, or voice-enabled products have a new competitor worth testing seriously. The Custom Voices suite and native document generation round out a platform that is maturing faster than most competitors expected.

Editor's Verdict

xAI Grok 4.3 Goes Public: $1.25 Per Million Tokens and Always-On Reasoning Shake Up the LLM Market earns a solid recommendation within the other llm space.

The strongest case for paying attention is most aggressive frontier model API pricing currently available ($1.25/$2.50 per million tokens), which raises the bar for what readers should now expect from peers in this space. Reinforcing that, 1M token context window suitable for enterprise-scale document analysis adds practical value rather than just headline appeal. The broader signal worth registering is straightforward: grok 4.3's $1.25 per million input tokens is approximately 75% cheaper than Claude Opus 4.7 and GPT-5.5, forcing immediate competitive repricing conversations across the industry. On the other side of the ledger, always-on reasoning increases latency, unsuitable for speed-critical or simple tasks is a real constraint, not a marketing footnote, and it should factor into any serious decision. Layered on top of that, higher per-token rates above 200K context limit the cost advantage for very-long-context workloads narrows the set of teams for whom this is an obvious yes.

For multi-model deployment teams, cost-conscious operators, and developers willing to evaluate beyond the major labs, this is a serious evaluation candidate, not just a curiosity to bookmark. For everyone else, the safer posture is to monitor coverage and revisit once the use cases that matter to your team are demonstrated in the wild.