ZH version is available. Content is displayed in original English for accuracy.
Advertisement
Advertisement
⚡ Community Insights
Discussion Sentiment
75% Positive
Analyzed from 1703 words in the discussion.
Trending Topics
#image#model#vision#models#deepseek#screenshots#images#https#text#flash
Discussion (66 Comments)Read Original on HackerNews
I asked it "what time does the clock show?" (both on reasoning: high)
DS answered: The clock shows *5:10* (and 45 seconds). Here is the breakdown: * *Hour hand (red, shortest):* Pointing at the *5*. * *Minute hand (green, longest):* Pointing at the *2*, which represents 10 minutes. * *Second hand (blue, medium):* Pointing at the *9*, which represents 45 seconds.
Qwen answered: The clock shows *8:10* (with the red second hand on the 5, i.e. *8:10:25*).
- *Hour hand* (short, blue) → 8 - *Minute hand* (long, green) → 2 (10 minutes) - *Second hand* (thin, red) → 5 (25 seconds)
Correct answer is 08:09:25.
> Images are converted into tokens based on their dimensions, and these tokens are billed together with your text tokens.
> Before inference, every image is automatically resized:
> - Images with a total pixel count below roughly 384×384 are scaled up while preserving their aspect ratio.
> - Larger images are scaled down while preserving their aspect ratio so that the total pixel count after resizing is roughly that of an 800×800 image.
> As a result, there is an upper bound of 384 tokens per image: for example, a 2000×2000 image and a 5000×5000 image consume the same number of tokens after resizing. When a request contains multiple images, each image is counted independently under the same rule—there is no separate calculation for multi-image requests.
400 tokens per image results in 2,500 images per dollar, if I’m not mistaken.
edit: format.
Maybe there are some use cases where you need high detail everywhere at once, but for OCR of small text and the like a zoom ability should be sufficient
Can't remember if I stole this idea from some existing public harness though, can't remember. If someone knows of public harnesses that do this already, please share them :)
Anecdotally, I had to tell 0731 to refrain from viewing screenshots since it kept breaking its sessions by trying to read images.
It's useful but for OCR and a lot of other applications it needs to be a bit higher (eg: putting in a full A4 / Letter sized page)
I am using a stripped-down minimal version of it which I uploaded here, since I am not a fan of huge dependency trees: https://github.com/99991/simple-pp-doclayoutv3
Another recent model for this task is Unlimited-OCR: https://github.com/baidu/Unlimited-OCR
https://openrouter.ai/deepseek/deepseek-v4-flash-vision-exp
I really missed this feature when I had DeepSeek code a small game for fun. When writing UI and rendering code it could execute the game and get screenshots back, but then had to rely on my feedback on what had gone wrong. Models with vision can do much better here, finding more issues on their own
Also used it for 3d printer control once, had it diagnosing issues, calibrating my Tradrack MMU and canceling failed prints autonomously from a couple of cameras placed around the printer.
It makes running much, much longer feedback loops possible. Although you can mix and match non-vision and vision models simply by invoking a vision model when you need one, as I like to use non-vision models like glm-5.3.
1. process graphs and charts
2. process handwritten math formula, also chinese characters writings
3. process design sketch and wireframe
4. process scanned documents
... etc
in fact these transformer models currently suck for surveillance, too slow and expensive. There are already faster and better facial/gait/object recognition models out there.
Edit: I see it has limited resolution. Luckily I just built a vision worker plugin for DSH that routes image input to Kimi K2.6 on Cloudflare.
This is useful for a reasonable amount of use-cases, but I think the watershed rez will be around triple that, ~1080p, which is enough for almost anything, except small text and subtle details.
But if not, does anybody know a recommended way to attach vision to deepseek flash (on a self-hosted infrastructure)?