Model Details
Grok-2 mini
Quick answer: Grok-2 mini (released August 13, 2024) is the fast, cost-efficient variant of Grok-2 — 10× cheaper at $0.20/$1 per million tokens, with the same 128K context window and real-time X data access. It was designed for high-volume use cases where Grok-2's full reasoning depth isn't needed for every query.
At a Glance
Where Grok-2 mini leads
- 10× cheaper than Grok-2: $0.20/$1 vs $2/$10 per million tokens
- Retains real-time X (Twitter) data access — the defining xAI advantage
- Fast response times for production-scale deployments
- Strong performance for its price tier: 70.1% MATH, 88.7% MMLU at launch
Where it lags
- Lower reasoning depth than Grok-2 for hard multi-step problems
- No native image generation (Aurora integration is Grok-2 full only)
- Superseded by Grok-3 Mini and later mini-tier models
Best for: high-volume chat, real-time X-integrated features, content classification, and any xAI workload where cost-per-call is the primary constraint.
What Grok-2 mini Is
Grok-2 mini launched simultaneously with Grok-2 as xAI's entry into the two-tier model strategy (full model + mini) that OpenAI and Anthropic had already established. The mini variant trades some reasoning depth for 10× lower cost and faster latency — making it accessible for high-volume applications like real-time X post analysis, automated moderation, and chatbot features.
Like Grok-2, it retains real-time access to X data — meaning it can answer questions about recent events, trending topics, and live discussions that knowledge-cutoff models miss.
Specifications
| Field | Value |
|---|---|
| Organization | xAI |
| Context window | 128,000 tokens |
| License | Proprietary |
| Release date | August 13, 2024 |
| Knowledge cutoff | June 2024 |
| Modality | Multimodal (text and vision) |
| Data access | Real-time X (Twitter) |
Pricing
| Input (per 1M tokens) | Output (per 1M tokens) | |
|---|---|---|
| xAI API | $0.20 | $1.00 |
Source: xAI API pricing.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| MATH | 70.1% | xAI — Grok-2 | 2024-08 |
| HumanEval | 88.7% | xAI — Grok-2 | 2024-08 |
| MMLU | 86.0% | xAI — Grok-2 | 2024-08 |
Scores reported by xAI. Not Benchgen measurements.
Grok-2 mini vs Alternatives
| Model | Context | Price (in/out per 1M) | Real-time data |
|---|---|---|---|
| Grok-2 mini | 128K | $0.20 / $1 | X (Twitter) |
| Grok-2 | 128K | $2 / $10 | X (Twitter) |
| GPT-4o mini | 128K | $0.15 / $0.60 | No |
| Claude 3.5 Haiku | 200K | $1 / $5 | No |
Grok-2 mini is priced comparably to GPT-4o mini with the unique advantage of real-time X data. Claude 3.5 Haiku offers better coding benchmarks and longer context at 5× higher cost. (Rival scores from respective announcements; not Benchgen measurements.)
Related Models & Benchmarks
Specs sourced from xAI — Grok-2. Last updated 2026-06-19.
Grok-2 mini
Grok-2 mini is a large language model developed by xAI.