# Grok‑2 mini

Grok‑2 mini is an AI model released by SpaceXSI on Aug 13 2024. Tracked results include 63.2% on MMMU.

## Facts

| Field | Value |
| --- | --- |
| Model | Grok‑2 mini |
| Developer | SpaceXSI |
| Release date | Tuesday, Aug 13 2024 |
| Licensing | Closed |

## Tracked benchmark scores

| Benchmark | Score | Source | What it measures |
| --- | --- | --- | --- |
| HumanEval | 85.7% | Lab | 164 small Python problems: the AI is given a function's description and has to write the working function. This was the coding benchmark of the GPT-3.5 and GPT-4 era, before the field moved to fixing real bugs in real repositories. Higher is better. |
| MATH | 73% | Lab | accuracy on the original competition mathematics test set |
| MMLU-Pro | 72% | Lab | accuracy on the ten-option MMLU-Pro benchmark |
| MMLU | 86.2% | Lab | A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. |
| MathVista | 68.1% | Lab | mathematical reasoning about visual information |
| MMMU | 63.2% | Lab | Tests the AI on understanding images and text together across many college subjects. Higher is better. |

## About Grok‑2 mini

Grok-2 mini entered beta on August 13, 2024 alongside Grok-2. xAI positioned the smaller model as a balance between response speed and answer quality, with access through Grok on X and enterprise API availability planned separately. It was a named sibling in the model family, giving users a lighter alternative to the larger Grok-2.

Its launch evaluations covered both text and visual understanding, including 72.0% on MMLU-Pro, 85.7% on HumanEval and 63.2% on MMMU. Those figures came from the mini model's own column in xAI's comparison, with HumanEval reported as pass@1. They should be distinguished from Grok-2's scores and from the separate image-generation model integrated into the Grok product.

## Questions and answers

### When was Grok‑2 mini released?

Grok‑2 mini was released by SpaceXSI on Tuesday, Aug 13 2024.

### Who made Grok‑2 mini?

Grok‑2 mini was built by SpaceXSI. Elon Musk's AI company building the Grok series of models. Founded in 2023 as xAI, merged into SpaceX as SpaceXAI in 2026 and renamed SpaceXSI (SpaceX Super Intelligence) that October. Cursor (Anysphere) and its Composer coding models were acquired in 2026 and are tracked here.

### What benchmark scores did Grok‑2 mini get?

Grok‑2 mini reports 6 tracked benchmark scores — HumanEval: 85.7%; MATH: 73%; MMLU-Pro: 72%; MMLU: 86.2%; MathVista: 68.1%; MMMU: 63.2%. Tracked scores may come from lab reports or independent benchmarks; source details accompany the benchmark data.

### Is Grok‑2 mini open source?

No. Grok‑2 mini is a closed model. The weights are not published — it is available only through the provider's own API, apps, or partner platforms.

### What came before and after Grok‑2 mini?

SpaceXSI's previous tracked release was Grok‑2 on Aug 13 2024. It was followed by Grok 3 on Feb 17 2025.


---

Canonical page: https://aireleasetracker.com/model/xai/grok‑2-mini
Site index: https://aireleasetracker.com/llms.txt
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.
