logoalt Hacker News

Inkling 2 Mini

174 points • by jasondavies • today at 8:12 PM • 19 comments • view on HN

Comments

montroser • today at 8:40 PM

This is bonkers good if these benchmarks hold up in real-world scenarios. To be at the level of Qwen3.8-27B with a looped transformer model is quite the flex. They just left this note on reddit:

Hope that folks find this model more amenable to local usage - the sizing was largely informed by what folks on this subreddit ask for the most haha!

There are two particularly interesting aspects of this model for local users:

We do our training end-to-end in nvfp4, and although we haven't exhaustively benchmarked this, folks like Unsloth have noticed that our models seem to respond very well to being quantized, maintaining 90% top-1 even at at 1-bit quantization.

This model is looped, which allows more depth at the same parameter count, which we particularly notice seems helpful for coding tasks. :)

This is our first model targeted at local users, so let us know how this model is and anything you want to see out of future iterations!

https://www.reddit.com/r/LocalLLaMA/comments/1y1diuk/thinkin...

yread • today at 9:01 PM

Recently SensaNova also released a looping model with unbelievable numbers

https://www.sensetime.com/en/news/sensenova-loop-9b-release-...

gchadwick • today at 8:47 PM

Very interesting to see them publish a looped transformer. It's been reported that OpenAI are now using looped transformers, though of course not confirmed.

I think this may be the first public example of a looped transformer model good enough for real world use cases?

rsolva • today at 8:58 PM

Going to test this right away on my AMD 7900XTX and see how it does compared to Qwen 3.8 27B (or DaisyRose 27B that I'm currently running). Given that Inkling 2 Mini is a MoE, it should provide way higher decode speed, but with comparable performance. Exiting!

jakswa • today at 8:44 PM

When inkling first released I tested it out and thought it was the best prose from an LLM in my quick sampling -- easy-reading and easy to understand. Big hopes I can write code with one of these models one day.

➕ show 1 reply
mattnewton • today at 8:21 PM

Benchmarks look good, and inkling 2 small felt better than the benchmarks suggested, so I am very excited to play with this.

From a research perspective, I'm even more interested in the varlooped variant they're releasing https://thinkingmachines.ai/model-card/tml-d24-varloop-resea... I'm not aware of any other working open source models like that.

➕ show 2 replies
neya • today at 8:32 PM

Wow, that coding demo was super impressive. Can't wait to try it out. Just need to wait for the hardware market to cool down a bit so we can run these on our own commodity hardware.

jakswa • today at 8:41 PM

The masses lying in wait for Qwen 4 cry out in joy. A savior? Could it be?!

walrus01 • today at 8:27 PM

Looking at the link to quantizations it looks like unsloth has already quantized it:

https://huggingface.co/thinkingmachines/Inkling-2-Mini

https://huggingface.co/unsloth/Inkling-2-Mini-GGUF

In terms of the active parameters and token/s on slow or less capable hardware it seems like it should be in the same general performance category as qwen 3.6 35B A3B.

➕ show 1 reply
christkv • today at 8:33 PM

I'm loving this trend. My framework desktop is looking like a better and better investment.

holistio • today at 8:46 PM

I'm starting to believe Gestalt is actually theirs. This is very nice.