Quick answer: Grok-2 (released August 13, 2024) was xAI's frontier model at launch, introducing native image generation via the Aurora model integration, enterprise API access at $2/$10 per million tokens, and significantly improved reasoning over Grok-1.5. It was the first Grok model available for commercial API use at scale.
Where Grok-2 leads
Where it lags
Best for: X-integrated applications, content generation with image output, and teams needing real-time social data alongside LLM capabilities.
Grok-2 was released at xAI's enterprise API launch in August 2024. It represented a major capability leap from Grok-1.5 — significantly better on GPQA (scientists-level questions), math, and coding, with the addition of vision understanding. The Aurora image generation integration made it the first commercial LLM API to offer combined text reasoning + image generation in a single provider relationship.
Grok-2 Mini launched simultaneously as a smaller, faster, cheaper variant for high-volume use cases where the full Grok-2 would be cost-prohibitive.
| Field | Value |
|---|---|
| Organization | xAI |
| Context window | 128,000 tokens |
| License | Proprietary |
| Release date | August 13, 2024 |
| Knowledge cutoff | June 2024 |
| Modality | Multimodal (text + vision input; text + image output via Aurora) |
| Data access | Real-time X (Twitter) |
| Input (per 1M tokens) | Output (per 1M tokens) | |
|---|---|---|
| xAI API | $2.00 | $10.00 |
Source: xAI API pricing.
| Benchmark | Score | Source | Date |
|---|---|---|---|
| GPQA | 56.0% | xAI — Grok-2 | 2024-08 |
| MATH | 76.1% | xAI — Grok-2 | 2024-08 |
| HumanEval | 92.7% | xAI — Grok-2 | 2024-08 |
| MMLU | 87.5% | xAI — Grok-2 | 2024-08 |
Scores reported by xAI. Not Benchgen measurements.
| Model | Context | MATH | Price (in/out per 1M) |
|---|---|---|---|
| Grok-2 | 128K | 76.1% | $2 / $10 |
| Grok-2 mini | 128K | 70.1% | $0.20 / $1 |
| Grok-4 | 256K | — | $3 / $15 |
| GPT-4o (Aug 2024) | 128K | 76.6% | $2.50 / $10 |
(Rival scores from respective announcements; not Benchgen measurements.)
Grok-2 is a large language model developed by xAI.