Full API Range: 4.3, 4.20, and Build 0.1

Updated November 2, 2026 — flagship model. Grok 4.5 is no longer the most capable model: Grok 4.6 (August 12, 2026) and then Grok 4.7 (grok-4.7, September 21, 2026) replaced it, with 500,000 context tokens, reasoning effort from low to xhigh, and the same tiered pricing ($2 / $6 under 200,000 prompt tokens, $4 / $12 above). grok-4.5 remains available; for a new project, start with grok-4.7 (course Grok in 2026, lesson 9).

The previous lesson introduced Grok 4.5; this one puts the flagship model in context, because xAI’s API catalog as of July 28, 2026, includes six text models, and the right choice rarely depends on the newest model alone. Let’s start with the official overview, including tiered context pricing, before outlining a selection method.

The catalog (docs.x.ai, recorded July 28, 2026)

(Price per million tokens.)

First observation, before even diving into model details: tiered pricing isn’t unique to Grok 4.5. All six models follow the same mechanics—the price changes around 200,000 context tokens—with different amounts. The habit of monitoring context sizes from the previous lesson therefore applies to the entire range.

How to read this range

At the top of the catalog, grok-4.5 represents maximum capability: code, agents, knowledge work, with configurable reasoning. It’s the model xAI highlights, and it’s the most expensive in the text catalog. Its 500k token window, comfortable in absolute terms, is still half that of the models following it in the table—we’ll come back to that in a moment.

Just below, grok-4.3 plays the role of the economical generalist with very large context. A one-million token window for $1.25 / $2.50 under the tier: this is the model for massive ingestion and corpus analysis. When the question is “How do I fit all this volume into a single request?”, this is the one—or the 4.20 variants—you should look at, not the flagship.

Next, understand the purpose of the three grok-4.20 variants, since they share exactly the same context and prices as 4.3. What they bring is fine-grained control over execution mode: the reasoning variant thinks before answering, the non-reasoning variant answers directly—faster, for tasks that don’t require deep reasoning—and the multi-agent variant orchestrates multiple agents. You therefore explicitly choose the behavior, where other models decide for you. This is the generation our existing courses detail, and it hasn’t disappeared with the arrival of 4.5.

Finally, grok-build-0.1 is a special case: it’s the fastest code model in the range, in public beta on the API since June 1, and the original engine behind the Grok Build agent. With its 256k context and floor pricing ($1 / $2 under the tier), it’s built for high-throughput code—use cases where you chain many requests and where latency and cost matter more than raw capacity.

How this changes your choices

The practical consequence of this catalog is that the question is no longer “Which Grok?” but “Which context/capability pair?”. If capability takes priority—demanding reasoning, agents, complex code—go with grok-4.5, watching the 200k tier that doubles the bill. If context takes priority, because you need to load a million tokens at once, grok-4.3 is the natural choice, or a 4.20 variant if you also need reasoning / multi-agent control. And if your need is high-throughput code at the best cost, grok-build-0.1 is made for that.

One final word on our existing courses: “Grok Model Guide” and “Multi-agent research” describe the 4.20 generation. Everything they say remains accurate—but this generation is no longer alone in the catalog. When you take them, simply place the 4.20 models in the table above: they occupy the middle position, between the flagship and the code specialist.

Official source: docs.x.ai/developers/models

Prices recorded August 5, 2026 — prices change regularly: before any budget calculation, check the current grid on the official xAI models and pricing page.

Key takeaways

• Six text models: 4.5 (500k), 4.3 and three 4.20 variants (1M), build-0.1 (256k)
• All follow tiered pricing around 200k context tokens
• The 4.20 variants retain the benefit of reasoning / non-reasoning / multi-agent control
• grok-build-0.1 is the fast, economical code model