DeepSeek Introduces Vision • Hacker News Board

Re: DeepSeek Introduces Vision

#141

Post by exabrial » Thu, Jun 18, 2026, 3:39 PM UTC

echelon wrote: post #131

exabrial wrote: post #128

The product I want most is the ability to return to the late January 2026 version of Anthropic models.

This is why we need open weights for everything. Nobody will cry when their AI girlfriend model gets revoked. You'll always have the weights. Presumably for the low cost of spinning up an H200 or two you can use the weights forever. No more claiming your LLM gets nerfed. No more claiming your video model can't do Spider-Man anymore.

I think my main concern was productivity, but tell me more about this AI Girlfriend

Re: DeepSeek Introduces Vision

#142

Post by dghlsakjg » Thu, Jun 18, 2026, 3:39 PM UTC

stijnveken wrote: post #129

ezconnect wrote: post #127

Earlier quoted context omitted.

What is Parakeet?

Maybe they meant narakeet? https://www.narakeet.com/tools/

Parakeet is the name of a speech to text model from Nvidia. Roughly comparable to whisper from openAI.

It's the model doing the work inside the wrapper that an app provides.

Re: DeepSeek Introduces Vision

#143

Post by codybontecou » Thu, Jun 18, 2026, 3:41 PM UTC

vitorgrs wrote: post #79

I already had it for months? What's the news here?

Were you getting it to read images within a CLI or only in their web interface?

Re: DeepSeek Introduces Vision

#144

Post by rob » Thu, Jun 18, 2026, 3:44 PM UTC

dghlsakjg wrote: post #142

stijnveken wrote: post #129

Earlier quoted context omitted.

Maybe they meant narakeet? https://www.narakeet.com/tools/

Parakeet is the name of a speech to text model from Nvidia. Roughly comparable to whisper from openAI. It's the model doing the work inside the wrapper that an app provides.

Yep, here's the v2 and v3:

https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2

https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3

It's almost instant on my new M5 Max w/ 36GB of memory, but I used both with Handy on my previous 2019 Intel Mac w/ 16GB memory and was completely surprised at just how fast it was for being on-device! Not instant, but only a couple seconds.

Re: DeepSeek Introduces Vision

#145

Post by superfrank » Thu, Jun 18, 2026, 3:46 PM UTC

k_138z wrote: post #73

I wonder what it has to say for the Tank Man image.

"It doesn't look like anything to me"

Re: DeepSeek Introduces Vision

#146

Post by carterschonwald » Thu, Jun 18, 2026, 3:59 PM UTC

anthonypasq wrote: post #120

testbjjl wrote: post #96

Earlier quoted context omitted.

DeepSeek interpreting screenshots and images I send it at fractions of what I pay Claude and ChatGPT, for me, is of far higher priority than supporting dictation. There are workarounds for dictation but not image processing.

just use one of the various cheap gemini models

gemini models are also fantastic at understanding non spoken sounds

Re: DeepSeek Introduces Vision

#147

Post by tehjoker » Thu, Jun 18, 2026, 4:25 PM UTC

bigmadshoe wrote: post #118

corimaith wrote: post #112

Earlier quoted context omitted.

Or you could just use a CNN...

CNNs are not SoTA anymore when it comes to large models, and also are not used to provide interpretations of images as text, but rather to classify, do semantic segmentation, etc.

Can you say more about that? I haven't kept up.

Re: DeepSeek Introduces Vision

#148

Post by fc417fc802 » Thu, Jun 18, 2026, 4:36 PM UTC

wolttam wrote: post #110

ForceBru wrote: post #89

Earlier quoted context omitted.

Wait, this isn't real, is it? Is there actually an intermediate model that translates DeepSeek's thinking from its "alien language" into human languages? That's not actually the case, right? I thought "thinking" is literally the model generating additional text in a human language that shows its "thought process". It's added to the model's context, which helps it reason better because it now has this self-generated c…

The parent is off, you’re right. They may reason in any language, typically whatever the user’s language is, and you’ll see the reasoning directly with an open model like Deepseek. Research only showed that thinking might be disconnected from the final output but in my experience they are very strongly correlated in recent models

> Research only showed that thinking might be disconnected from the final output

It is trivial to regularly spot obvious contradictions and inconsistencies if you read carefully. For example I've encountered traces that amounted to "I can deduce X, therefore Y, so that means Z" but then the model turns around and outputs "the answer is W because X". It's even been demonstrated that having the model output placeholder tokens or other gibberish instead of "thoughts" still improves performance. However the thinking traces can still be useful to the end user regardless.

Re: DeepSeek Introduces Vision

#149

Post by holoduke » Thu, Jun 18, 2026, 4:39 PM UTC

A bit of topic. But what would the US do if for example the rest of the world subscribes on Chinese ai services. I think the US would show some really nasty behavior.

Re: DeepSeek Introduces Vision

#150

Post by fc417fc802 » Thu, Jun 18, 2026, 4:39 PM UTC

ForceBru wrote: post #89

cocoflunchy wrote: post #43

Earlier quoted context omitted.

I believe that most reasoning models actually think in their own "language" which is not really understandable by humans. The thinking traces that are shown in the UI are actually summaries generated by a smaller model in plain english (or user language). Sometimes this leaks through and you see some chinese/japanese characters in e.g. Claude's reasoning.

Wait, this isn't real, is it? Is there actually an intermediate model that translates DeepSeek's thinking from its "alien language" into human languages? That's not actually the case, right? I thought "thinking" is literally the model generating additional text in a human language that shows its "thought process". It's added to the model's context, which helps it reason better because it now has this self-generated c…

Current models simply generate additional text that gets added to the context for the trace. However iterative models that "think" by repeatedly looping through several layers instead of outputting text have recently been demonstrated.