XCSme

9 ms·

241.
by XCSme 4mo ago
I remember when I was learning WebGL/OpenGL and had to draw some test shapes to test my shader, I would manually think what the vertex positions should be to draw a triangle, a pyramid, etc. I think for AIs it's quite easy, becaus
242.
by XCSme 4mo ago
I kind of disagree, because describing an image is more like writing an svg than rendering pixels. If I ask you to describe how would you draw a cat head, you could do it in text like: "a big circle as the head, 2 small circles as the
243.
by XCSme 4mo ago
I use V4 Flash as my main model, it really is exceptionally capable for the price. It can follow instructions quite well, if you have a plan, it usually executes it to completion and you get code that works. My only concern with it, is that
244.
by XCSme 4mo ago
Well, it is US English I think, I could add UK English too.
245.
by XCSme 4mo ago
Svgs are just shapes, they (models) mostly only use sqaures, circles, etc. Rarely paths. Plus most models are multi-modal, so they have a visual understanding too.
246.
by XCSme 4mo ago
Sorry, accidentally posted the Romanian version, lol, here is en: https://aibenchy.com/showcase/hamster-playing-table-tennis-s... If mods could update the link please.
247.
SVG of a Hamster Playing Table-Tennis(aibenchy.com)
21 points by XCSme 4mo ago|18 comments
248.
by XCSme 4mo ago
Not if the original code is secure...
249.
by XCSme 4mo ago
Any reason why they are not on openrouter yet, but the speech models are? https://openrouter.ai/models Curious to test them and see how they perform.
250.
by XCSme 4mo ago
Slighty weaker than Qwen 3.7 Max, but 2x cheaper on my tests[0]. [0]: https://aibenchy.com/compare/qwen-qwen3-7-max-medium/qwen-qw...
251.
by XCSme 4mo ago
Not sure if considered it's considered small in any way, but DeepSeek V4 Flash is really decent.
252.
by XCSme 5mo ago
A lot better than their previous models: https://aibenchy.com/compare/minimax-minimax-m2-7-medium/min...
253.
by XCSme 5mo ago
I will test it when it's accessible via OpenRouter, but the previous LFM2 model (lfm-2-24b-a2b) didn't do well on my tests, it got only 1/20 questions/tasks right, way below Gemma 31B or Qwen 35b-a3b (those get like 10&#
254.
by XCSme 5mo ago
lol, same result but without the explanation on why: https://gemini.google.com/share/96f93fe601c6
255.
by XCSme 5mo ago
Also, what about the major flaw/bias linked for Gemini 3.5 flash? That has major real-life consequences if the model ends up being used for any automated scoring systems. I found it while trying to use 3.5 Flash for scoring the reasoni
256.
by XCSme 5mo ago
I'm happy you do comment, I did add more coding tests since then and add more improvements (price history per model, displaying cost to run at current pricing, improved scoring). How is it useless to see that Opus 4.8 is 2x more expens
257.
by XCSme 5mo ago
Also check mine[0], basically random private tests/questions and an ok-ish methodology, testing mostly for general intelligence than coding-specific tasks. I built it for myself, to test which models to use via OpenRouter for my n8n ag
258.
by XCSme 5mo ago
The total cost of running my benchmarks, was 1.6x higher compared to Opus 4.7, mostly because of 2x output tokens: https://i.snipboard.io/vrdwTa.jpg
259.
by XCSme 5mo ago
For some reason everything is 2x (2x cost, 2x avg response time, 2x reasoning and output tokens)... Double-checking my test harness, but it's the first model that does this, so I doubt the issue is on my side... EDIT: Harness seems cor
260.
by XCSme 5mo ago
On my tests[0] it does a bit worse, and it's almost 2x expensive than Opus 4.7... I was surprised to see that it failed a Data extraction test (it gets it right 2/3 times, but one time it randomly returns null for a value instead)
261.
by XCSme 5mo ago
In some countries with good pro-tenant laws, renting might be "safer", as you have more protection if anything goes wrong. If you can pay the rent, they can't really legally kick you out. It something goes wrong with the apt.
262.
by XCSme 5mo ago
It seems to be back up now...
263.
by XCSme 5mo ago
We can no longer make fun of the one 9 of availability if we can't see it... The linked incident is related to this Copilot issue, where people lost access to most models: https://www.reddit.com/r/GithubCopilot&#x
264.
by XCSme 5mo ago
Work same amount of hours -> more productive -> client happier
265.
by XCSme 5mo ago
Nice, at what scale?
266.
by XCSme 5mo ago
Is anyone here using Deno in production?
267.
by XCSme 5mo ago
To save you a click, this is the output: Evaluation Based on the final line (Result: 3,5,7) and the provided grading criteria, here is the compressed evaluation: Rating: 7/10 Rationale The final line explicitly
268.
by XCSme 5mo ago
Direct link to the chat, ignore the story, it's just some filler tokens: https://gemini.google.com/share/244af1e74841
269.
Tell HN: Gemini 3.5 Flash breaks in stupid ways
9 points by XCSme 5mo ago|4 comments
270.
by XCSme 5mo ago
Well, won't it be hard to test? The customers could be paying for "increased visibility" in LLM results, but nothing specific. Also, the ad platform could simply hint towards a customer, instead of directly mentioning, thus n

More ›