← Back to directory
xAI
Grok 3
Released Feb 17, 2025 · knowledge cutoff 2024-11
- Status
- Superseded
- Location
- United States
- Modality
- Multimodal
- Context window
- 131K
- Max output
- 16K
- Speed
- —
- Price ($/MTok in / out)
- $3 / $15
- Cost per task
- —
- Open weights
- No
Benchmarks4 of 8 reported
MMLU-Pro
79.9%
GPQA Diamond
84.6%
AIME (latest)
93.3%
LMArena Elo
1402
SWE-bench Verified
—
Terminal-Bench 2.1
—
Humanity's Last Exam
—
ARC-AGI-2
—
Strengths & weaknesses
Strengths
- Pulls live posts and breaking events into answers instead of relying on training data alone
- Blunt, unhedged register — it commits to an answer where others disclaim
- Reasoning mode available on demand for maths and puzzles
Weaknesses
- Confident on live-data questions even when the underlying posts are wrong
- Uneven at code compared with its maths
- Reachable only through the X subscription tiers at launch
Reasoning-mode (beta) scores shown.
For developers
API model strings
Not researched
Licence
Proprietary — weights not released
Retirement
No retirement announced
Lineage
No recorded predecessor