# H2O AI Personal GPT — user reviews

> 15 indexed App Store user reviews of H2O AI Personal GPT by H2o.ai Inc, rated 2.80/5 from 32 ratings.

- Source: https://appshunter.io/ios/app/h2o-ai-personal-gpt/id6504365990/reviews
- App overview in markdown: https://appshunter.io/ios/app/h2o-ai-personal-gpt/id6504365990.md
- Rating breakdown: 5★ 13 · 4★ 0 · 3★ 2 · 2★ 1 · 1★ 16

## What do users think of H2O AI Personal GPT?

**Pros**

- 👍 Decent while in airplane mode, no network needed, still gives sensible answers
- 👍 Good data privacy practices
- 👍 Able to answer correctly some specific questions

**Cons**

- 👎 Weak intelligence, often gives non-sensical responses
- 👎 Some models hallucinate terribly
- 👎 Inconsistent performance, may loop in repetitive answers

## Most critical reviews (10 lowest-rated)

### 1/5 — 闪退

*2024-11-16, version 1.0.7*

更新后，任意页面，返回操作会闪退

### 1/5 — 体验使用

*2024-09-04, version 1.0.4*

没有任何作用，答非所问，1.5GB的大小。

### 1/5 — 一段话或一个字一直重复

*2024-07-30, version 1.0.3*

…

### 1/5 — 太垃圾了！

*2024-07-29, version 1.0.3*

简直就是弱智！

### 1/5 — Stupid

*2024-07-18, version 1.0.3*

Stupid enough to make me cry

### 1/5 — 手机发热严重 怀疑程序有问题

*2024-07-17, version 1.0.3*

手机发热很严重了

### 1/5 — 失败的ai语言模型

*2024-07-16, version 1.0.2*

虽说是给予gpt 但语言逻辑不通，检索都是瘸腿状态，所言非所答，完全无法使用。请不要下载 ，这在严重浪费你的时间。

### 1/5 — 不多评价了

*2024-07-16*

人工智障…..

### 2/5 — Doesnt work?

*2025-07-12, version 1.0.13*

I give it a prompt and it does not respond. I exit the chat to view other chats and the app just crashes entirely. Wish it worked :/

### 2/5 — Promising models, but not as useful due to major glitch

*2024-09-22*

The Danube models are quite promising given how smoothly they run on an iPhone and the quality of output given their compact size!

Unfortunately the app itself isn’t proving as useful due to the fact than trying to paste any text longer than a paragraph or two into the chat results with a crash.

If you’re trying to get the model to interact with content such as a short article (e.g. to test its summarisation skills) or a longer list of points from a note (e.g. to test its capability to generate prose), or you simply have a library of more rich prompts you use to test and compare models, then this app is not going to work due to that glitch.

@H2O team: ♥️ your effort on Danube models! Please consider fixing this problem to make users on mobile devices enjoy them at their maximum capability.

## Most praised reviews (5 highest-rated)

### 5/5 — 大便一样的智商

*2024-08-08, version 1.0.4*

浪费流量。弱智产品

### 5/5 — Good for what it’s supposed to do

*2024-07-31, version 1.0.3*

I think this is an extremely interesting example of AI/LLMs on the edge. No network connection, meaning this works while in airplane mode, and it still gives generally sensible answers for most prompts. As an example of good data science practices (data security and privacy) along with being about as good as network connected GPTs, I’d say this is a winner

### 5/5 — Danube-3 4B is decent

*2024-07-23, version 1.0.3*

Though smaller models are just laughable, Danube-3 4B is decent. When I asked who are Jobs and Cook, it was able to answer correctly. While the smaller models hallucinate terribly

### 3/5 — 出错的概率比较大

*2024-08-11, version 1.0.4*

我想用它来做翻译，但是他翻译的水平实在是不敢恭维。

### 3/5 — 离线模式可用

*2024-08-06, version 1.0.3*

值得鼓励，使用起来几乎是零门槛，必须点赞！但它时常发神经，不断死循环重复几句话；回答的科学方面的问题，90%以上是错误的（在线状态）。希望更新之后的版本越来越好

---

*Data collected daily from the US App Store and indexed by [AppsHunter](https://appshunter.io/). User reviews are verbatim App Store reviews. Ratings, prices and chart positions refresh continuously; this snapshot is from 2026-07-08.*