Million-token context
Suited to keeping long documents, large repo slices, and multi-turn history in one thread.
xAI · Grok 4.3 · Apr 2026
Grok 4.3 — a long-document understanding model with million-token context
- Million-token context
- Adjustable reasoning depth
- Tool calls that follow instructions
- Structured output
What is Grok 4.3?
Grok 4.3 is xAI’s model from around April 2026. It offers million-token context and is most often used to finish long documents or large repo slices in one thread—then follow instructions for tool calls and structured output. Versus earlier Grok 4 family routing, 4.3 is the general long-context pick; when multi-step coding, retrieval, and code execution matter more, compare Grok 4.5. On iMini Agent, pick Grok 4.3 to use it.
- Vendor
- xAI
- Released
- Apr 2026
- Context
- 1M tokens
- Max output
- Per platform limit
- Deep reasoning
- Adjustable · can run lighter day to day
- Best for
- Long docs & instruction-following tool calls
Where it stands out
Long context, tool calls, and adjustable depth—the boundaries to check before you stay on 4.3.
Adjustable reasoning depth
Switch from lighter to deeper reasoning by task; everyday work can run lower depth for speed.
Tools and structured output
Function calling and structured output hold up better—suited to business systems and automation flows.
More restrained claims
xAI’s cloud listing materials stress relatively lower hallucination—suited to jobs that need clear conclusions with less filler.
Three common workflows
Grok 4.3 fits long-material reading and tool calls that must follow instructions tightly; for heavier multi-step coding, compare 4.5.
Long-document analysis
For reading and summarizing contracts, manuals, and multi-file materials. Anchor claims to the source text.
Instruction-following tool calls
For business flows that demand strict instruction following. Lock success criteria and an output schema.
Q&A and medium coding
For Q&A, planning, and medium-difficulty coding. Step up to 4.5 when coding plus retrieval/execution gets heavier.
How to choose vs Grok 4.5 and Gemini 3.1 Pro
All three are modern primary picks. The gap is mainly context length, coding tool surface, and product stack.
| Criterion | Grok 4.3 | Grok 4.5 | Gemini 3.1 Pro |
|---|---|---|---|
| Context | 1M tokens | 500K tokens | 1M tokens |
| Reasoning | Adjustable depth · can run lighter day to day | Leaner tokens · tools ready | Adjustable thinking depth |
| Coding & agents | Long materials & medium coding | Multi-step coding with retrieval & execution | Multimodal heavy coding |
| Long-horizon tools | More headroom in long threads | Search & code execution ready | More complete multimodal orchestration |
| Speed posture | Can trade depth for speed | Favors tool-collaboration efficiency | Favors precision |
| Prefer when | Long-doc jobs that need million-token context | Heavier coding, retrieval & code execution | Hard work on the Google stack |
Using Grok 4.3 on iMini
Free quota, model naming, key specs, and how it differs from 4.5.
Use Grok 4.3 free on iMini Agent
Million-token context—suited to long-material reading and instruction-following tool calls.
Open Grok 4.3No xAI API key required.