← Back to directory
xAI
Grok 4.3
Released Apr 30, 2026 · knowledge cutoff unpublished
- Status
- Unknown
- Location
- United States
- Type
- LLM
- Modality
- Multimodal
- Context window
- 1M
- Max output
- —
- Speed (high effort)
- 117 tok/s · 17s to first answer token
- Price ($/MTok in / out)
- $1.25 / $2.50
- Cost per task (high effort)
- $0.14
- Open weights
- No
Benchmarks1 of 10 reported
GPQA Diamond
90.1%
MMLU-Proretired
—
SWE-bench Verified
—
SWE-bench Pro
—
Terminal-Bench 2.1
—
AIME (latest)retired
—
Humanity's Last Exam
—
LMArena Elo
—
GDPval-AA v2
—
ARC-AGI-2
—
Strengths & weaknesses
Strengths
- Very high output throughput once it starts — fast for a model that always reasons
- Takes video files directly, with no frame extraction or transcription step
- Reasoning is always on, so there is no mode to pick and get wrong
Weaknesses
- Long wait before the first token, because it always thinks first
- Still states wrong facts confidently enough to need grounding for factual work
Requests above 200K tokens are billed at 2x. GPQA Diamond 90.1 per Artificial Analysis (third-party). xAI documents no fixed max-output cap; xAI published agentic (tau2-Bench, IFBench) rather than academic benchmarks at launch.
For developers
API model strings
Not researched
Licence
Proprietary — weights not released
Retirement
No retirement announced
Lineage
No recorded predecessor