On 19 September 2025 xAI released two models labelled Grok-4 Fast:
grok-4-fast-reasoninggrok-4-fast-non-reasoning
According to the official post, both versions offer a 2 million token context window and are advertised as “up to 10× faster” than the standard Grok-4 model.
| Item | Quoted figure | Source |
|---|---|---|
| Context window | 2 000 000 tokens | xAI, 19 Sep 2025 |
| Speed-up vs Grok-4 | “up to 10× faster” | xAI, 19 Sep 2025 |
| API input price | $0.20 per 1 M tokens | console.x.ai pricing page |
| API output price | $0.50 per 1 M tokens | console.x.ai pricing page |
“Grok-4 Fast sets a new record on the intelligence-cost Pareto frontier.”
— Artificial Analysis, 20 Sep 2025
LMSYS reported the following arena scores (20 Sep 2025):
- Search Arena: 1342 Elo (rank 1)
- Chat Arena (text only): 1287 Elo (rank 8)
The models are accessible at grok.com and through the xAI API.
Join the discussion. You need to be logged in.