# \[ANN\] ollama-haskell-0.2.1.0 released

**URL:** <https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024>\
**Category:** Announcements\
**Created:** [September 24, 2025, 1:27pm UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024 "2025-09-24T13:27:16Z")\
**Posts on this page:** 8\
**Page:** 1

<div class="post-metadata">

**Author:** ![tusharad](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.haskell.org/tusharad/32/5133_2.png) [@tusharad](https://discourse.haskell.org/u/tusharad)\
**Post date:** [September 24, 2025, 1:27pm UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024/1 "2025-09-24T13:27:16Z")

</div>

I’m happy to announce the release of **ollama-haskell-0.2.1.0** , a Haskell client library for interacting with [Ollama](https://ollama.com/).

This release includes:

1. Compatibility with **Ollama 0.12.1**

2. Support for the **Create Blob** option

3. Standardized field names across modules for consistency

4. Ability to specify **dimensions options for embeddings**

📦 Hackage: [ollama-haskell](https://hackage.haskell.org/package/ollama-haskell)  
💻 GitHub: [tusharad/ollama-haskell](https://github.com/tusharad/ollama-haskell)

As always, feedback, bug reports, and contributions are very welcome!

---

<div class="post-metadata">

**Author:** ![peregrinum](https://avatars.discourse-cdn.com/v4/letter/p/d26b3c/32.png) [@peregrinum](https://discourse.haskell.org/u/peregrinum)\
**Post date:** [September 24, 2025, 6:07pm UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024/2 "2025-09-24T18:07:31Z")

</div>

This is really cool, definitely going to try it out.

It’s been a while since it played around with ollama, how does it compare now to other models like Claude Gemini etc? And for using it on Haskell code bases?

---

<div class="post-metadata">

**Author:** ![tusharad](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.haskell.org/tusharad/32/5133_2.png) [@tusharad](https://discourse.haskell.org/u/tusharad)\
**Post date:** [September 24, 2025, 6:39pm UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024/3 "2025-09-24T18:39:43Z")

</div>

Every week, some new model comes out that exceeds expectations. I personally love using the Qwen3 model. These models are obviously not as smart as Gemini or Claude, but it feels good not to have to worry about exhausting my API key!

---

<div class="post-metadata">

**Author:** ![asjo](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.haskell.org/asjo/32/294_2.png) [@asjo](https://discourse.haskell.org/u/asjo)\
**Post date:** [September 26, 2025, 8:20am UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024/4 "2025-09-26T08:20:01Z")

</div>

The link to [examples/OllamaExamples.hs](https://hackage.haskell.org/package/ollama-haskell-0.2.1.0/examples/OllamaExamples.hs) in the Readme on Hackage gives me a Page Not found.

The same happens on GitHub: [examples/OllamaExamples.hs](https://github.com/tusharad/ollama-haskell/blob/main/examples/OllamaExamples.hs): “404 - page not found The main branch of ollama-haskell does not contain the path examples/OllamaExamples.hs.”

I guess the link was supposed to go to the folder `examples/ollama-example/`?

Two thumbs up for having full examples! I’m especially interested in tools/MCP.

---

<div class="post-metadata">

**Author:** ![mimi1vx](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.haskell.org/mimi1vx/32/5053_2.png) [@mimi1vx](https://discourse.haskell.org/u/mimi1vx)\
**Post date:** [September 28, 2025, 4:52pm UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024/5 "2025-09-28T16:52:59Z")

</div>

ollama isnt model but model runner/manager 😃 So question how it compares with other models is pointless.

Better question is how it compares to LM Studio/LocalAI/llama.cpp,

what is important Ollama 0.11 added support for “cloud” models behind paywall so you can use your own local instance or if you need model with big hw req. you can pay for it run in the cloud.

---

<div class="post-metadata">

**Author:** ![chrisdone](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.haskell.org/chrisdone/32/1408_2.png) [@chrisdone](https://discourse.haskell.org/u/chrisdone)\
**Post date:** [September 29, 2025, 9:26pm UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024/6 "2025-09-29T21:26:52Z")

</div>

> [@mimi1vx](#):
>
> Better question is how it compares to LM Studio/LocalAI/llama.cpp,

I’ve dabbled a modest amount with ollama and llama.cpp (written up in this diary: [LLMs](https://chrisdone.com/posts/llms/) — see heading 2025-08-19), the notable difference being **GBNF support** for me personally. I haven’t explored the others.

---

<div class="post-metadata">

**Author:** ![mimi1vx](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.haskell.org/mimi1vx/32/5053_2.png) [@mimi1vx](https://discourse.haskell.org/u/mimi1vx)\
**Post date:** [September 30, 2025, 8:55pm UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024/7 "2025-09-30T20:55:53Z")

</div>

read your blog and this is typo or something wrong in your setup? `I’ve experimented with local LLM models on my MacBookPro Max M4, which can run 3b parameter models.` ?  
Because on my mini M4 I can run models with 30b parameters ..

---

<div class="post-metadata">

**Author:** ![chrisdone](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.haskell.org/chrisdone/32/1408_2.png) [@chrisdone](https://discourse.haskell.org/u/chrisdone)\
**Post date:** [October 1, 2025, 9:24pm UTC](https://discourse.haskell.org/t/ann-ollama-haskell-0-2-1-0-released/13024/8 "2025-10-01T21:24:02Z")

</div>

Yep, I’ve been experimenting with e.g llama3.2:3b and similar. I found that larger models are slower than I’d like in my MacBook, whereas the 3b ones are snappy.