Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Inkling: Our Open-Weights Model

thinkingmachines.ai

Post reply on HN
Print view View original

291–300 of 324 posts

Re: Inkling: Our Open-Weights Model

#291

Post by smiths1999 »

swiftcoder wrote: post #249
throwaw12 wrote: post #234

Earlier quoted context omitted.

What does it mean, if it is American? Is it censored or will it eventually stop working in Middle Eastern countries? Or is it biased towards powerful political lobby group interests? When weights are open I usually don't care where is it from, as long as it is working for my use cases well

There's a lot of Sinophobia in the tech world, and the AI race seems to be magnifying that. With very little evidence to back it up, I might add - we've had regular releases of quality open-weight models from Chinese firms, and no sign they are more censored/ideological, than, say, Grok

If the company does any kind of work with the US Government it's far easier to just say no to Chinese-origin models rather than deal with all the overhead on the government contracts. Since just about every major tech company has US contracts they all have to be careful about the use of Chinese models.

Re: Inkling: Our Open-Weights Model

#293

Post by alightsoul »

Certhas wrote: post #267
alightsoul wrote: post #225

Earlier quoted context omitted.

Ask an American model about the Epstein files and trump's involvement of them. Or about Israel committing genocide in Palestine. Every model will have their own bias. Freedom is relative and American freedom is not the only form of freedom. China bad or China authoritarian and thus not free is a result of the red scare

Here is what Sonnet says on genocide: "Surveys of genocide scholars (e.g., one by the journal Journal of Genocide Research contributors) show meaningful disagreement, though a visible and growing share of specialists in genocide studies specifically have concluded the term applies or is defensible — more so than international law scholars generally, who tend to be more cautious about the intent requirement." You can…

You asked about genocide in general. Not Israeli genocide. Are you astroturfing? No one can make an omission that obvious

You can't genuinely argue that the US, for all its problems, is less free than China

This is how Americans convince themselves they are at the top of the world when they are not. No one outside of Europe and the US cares about the US anymore. They buy Chinese products and do business with Chinese businesses instead, right now. Worldwide, China is seen more positively than the US specially in global south countries. The exceptions of course are Europe and The US, the cold war first world countries of course.

freedom house index

This is obvious American bias. Astroturfing.

Re: Inkling: Our Open-Weights Model

#294

Post by ekianjo »

paxys wrote: post #119
segmondy wrote: post #65

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.…

Not to mention - it is American. This is the first competitive non-Chinese open weights model since what, Llama 3?

gpt oss was competitive when it cam out

Re: Inkling: Our Open-Weights Model

#295

Post by ascent817 »

mark_l_watson wrote: post #286
ascent817 wrote: post #189

Earlier quoted context omitted.

There's also poolside.ai

I was blown away by how effective their latest model drop that works with their own coding harness pool is. I tested it extensively with Haskell, Python, and TypeScript for small coding projects. The functionality is good but the inference speed is too slow for most of my work: I would set up a problem, take a walk, then return later to evaluate the results. Note that I have an old mac mini with 32B memory; a fast mo…

Yeah, they’re focused on selling to government and defense contractors. I first encountered poolside at my aerospace internship.

Re: Inkling: Our Open-Weights Model

#296

Post by nullbio »

segmondy wrote: post #65

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.…

I'm sure it's better than KimiK2.7 and GLM5.2. Benchmarks aren't the full picture. Despite GLM5.2 performing well on benches and supposedly near frontier, in reality it was nothing close to frontier in actual usage.

Re: Inkling: Our Open-Weights Model

#297

Post by nullbio »

Topfi wrote: post #102

Very preliminary testing so far, but there is something here, far beyond what the benchmarks suggest. Only ever saw such outperformance of public evals vs my private ones with Anthropic models and while it is far to early to make any judgement at this stage, this model will take up a lot of mine time in the coming weeks by the look of things. Only ever viewed Moonshot AIs models as something I'd be able to live with…

[dead]

Re: Inkling: Our Open-Weights Model

#298

Post by altmanaltman »

hattmall wrote: post #197
altmanaltman wrote: post #154

Earlier quoted context omitted.

https://chat.deepseek.com/share/uf8jih4q95lbq3s8t2 This is what Deepseek replied when I asked it with a burner account. Claims it doesn't have it in its training data... sure.

Be interesting to see if it's possible to jailbreak and have it talk about it.

I am sure it can be but like it does prove the point of state interference in what the LLM is allowed to say/do in this case.

Re: Inkling: Our Open-Weights Model

#299

Post by dipakk »

is there something in the space of "taskifying" enterprises data for them? inkling on its own looks high-quality, but expecting companies to spend $ figuring it out before spending more $ on the actual fine-tuning job seems ... hard, especially if making the model especially customizable is the goal? or do the unit economics just work out with a small number of fine-tuners training a ~1T model on Tinker?

Re: Inkling: Our Open-Weights Model

#300

Post by hibijibies »

hham wrote: post #273
aabhay wrote: post #104

What strikes me the most is just how many different tasks are involved in modern model design. It used to be the case that you come up with a new loss function, slight architecture changes, etc., run your train and eval loop, and publish the artifacts. Now, there’s so much work to do just to keep up. It’s the ultimate red queen race. All of the 500 steps involved, each of which is its own little optimization loop, is…

https://news.ycombinator.com/item?id=48892559

It's probably not that many people necessary for creating a decent model. Soofi was small team: https://news.ycombinator.com/item?id=48870978
Post reply on HN