Inkling: Our Open-Weights Model

219 points65 comments2 hours ago
segmondy

Very nice, multi modal, largest open weight model that supports audio. Would be interesting to see how good the audio capability is.

If you want to run locally, checkout https://github.com/danielhanchen/llama.cpp/tree/add-inkling https://unsloth.ai/docs/models/inkling https://huggingface.co/unsloth/inkling-GGUF https://huggingface.co/unsloth/inkling-NVFP4

This supposedly is better than KimiK2.7, as much hype as GLM5.2 gets, I find myself using KimiK2.7 half of the time, so if the benchmark is true, then this can definitely go in the mix. My hope is that it might have strengths in some areas to beat all other open weight models.

ls_stats

America needs its own DeepSeek or Z.ai, a lot of people (myself included) root for open chinese models to win because they have no other choice.

Thinking Machines might be it.

show comments
ianbutler

It's nice to see a strong long context open weights model that is multi-modal.

There are many applications that will benefit from the strength in audio here and until z.ai and co work in visual this could be very strong for general agentic applications, though I see there's a bit of weakness in the benches for areas that might make that less true.

Like all models need to slap it in your harness and do proper evals on the tasks you care about.

show comments
minraws

For a first model, and given it's open, I am gaining some faith in American Open research labs again...

I couldn't test it since it's not on openrouter or something, but even if it's only as good as GLM5.1 that's more than good enough first attempt, I think.

Perhaps a lot more labs will catch up to ballpark frontier esque level soon, I am all for more competition in any field.

janalsncm

For the most part it’s better than Nemotron, worse than GLM. This makes it the best American open weights model from what I can tell?

show comments
Reubend

Seems like this is particularly good at instruction following, but not as strong at coding as others. It's always great to get more diversity of open weight models though! I'll need to test this out to see what its "personality" is like.

dr_dshiv

What are the different business models for open-weight AI companies?

show comments
GodelNumbering

Interestingly, when opening this page, the first thought I had was not that the benchmarks should be high, but 'I really hope they did not benchmaxx'. I think a model with modest benchmark scores can have much better real world utility as opposed to the current frontiers that are RL'd into being robotic and rigid.

firasd

Looks like it can be tried at https://tinker.thinkingmachines.ai/playground

alansaber

I never thought i'd see the day they released a model, rather than a blog post. The Figure 3 demo being a screencap of chrome in localhost made me feel better about myself. Jokes aside, best western open weights model- very cool.

show comments
ggcr

My personal bet is that this model should really shine in Autoresearch NanoGPT-style speedruns because its first-class integration with Tinker

bbstats

too bad we'll never know how good it is, since they used a radar plot to show its benchmark scores!

show comments
mhluongo

Interested in the implied strategy - that training a bespoke model for what you need will make economic sense over using a mass-trained model. I wonder if that's true?

androiddrew

Give me a good 180B param model that fits snuggly on an single DGX spark and I will sing your praises.

pants2

The Artifical Analysis has a link on their homepage but it 404's :/

https://artificialanalysis.ai/models/inkling

trilogic

You certainly cooking smth, Good Luck Mira.

amarble

They also indicate they have a 276B A12B version, but it doesn't seem the weights are available. This might actually be able to fit in 128GB when quantized to 2 bits or so which makes it interesting.

show comments
RohoSwagger

why is this website ai slop

solomatov

It looks like HuggingFace shows Apache-2.0 but they have AUP. How does it work together?

bobkb

Happy to see an open weight model ! This has all the right ingredients for success.

inkvi

Do they have an api to try the model in real envs?

verdverm

If it's ~30% bigger and not as good as GLM 5.2, why would I tinker with this model?

Maybe for the multi modal?

show comments
MaxPock

Raised 2 billion dollars at a 12 billion valuation and debuts at 41 on the Artificial Analysis Intelligence Index, while KIMI and DeepSeek will release Fable-class models this week. What a joke.

show comments
raverbashing

Cool, now we just need the GPU that supports it