News for hackers. Stuff that matters.
Re:Perplexity Decider v1.3: Training, eval, and use cases by redanddead on Friday November 13, 2026 @03:19PM (#50489067)
Complete lack of creativity in the field of AI...

Here's new product, which is just our shitty version of this easily reproducible technology

What is this?

Re:Perplexity Decider v1.3: Training, eval, and use cases by jgalt212 on Friday November 13, 2026 @03:25PM (#50489144)
Perhaps, but if you're already inside GCP it's nice to have something easy to turn on.
Re:Perplexity Decider v1.3: Training, eval, and use cases by adrianvi on Friday November 13, 2026 @03:57PM (#50489557)
Not just that, if your company approved and gets billed by X cloud, it's quite difficult to request another product from Y cloud
Re:Perplexity Decider v1.3: Training, eval, and use cases by agentdev001 on Friday November 13, 2026 @03:27PM (#50489172)
You seem to be confusing market capture for novelty. This almost certainly is not intended to be taken as some greenfield thing from perplexity, same goes for google
Re:Perplexity Decider v1.3: Training, eval, and use cases by happyopossum on Friday November 13, 2026 @03:52PM (#50489476)
[delayed]
Re:Perplexity Decider v1.3: Training, eval, and use cases by sowhat1 on Friday November 13, 2026 @05:39PM (#50491098)
And all the people involved are making millions to billions. Maybe we're the schmucks.
Re:Perplexity Decider v1.3: Training, eval, and use cases by waximabbax on Friday November 13, 2026 @03:20PM (#50489087)
Well congrats.

Here are some thoughts not facts:

On AA Decision Index.

Decider v1.3 and Decider v.1.2 seem to have same benchmarks. However, V1.3 flash beats V1.2 flash by 4 points.

I am not particulary excited about this pattern/trend of having flash versions for decision models. The whole point of Decision models was to deliver faster ouput for high volume requests at much cheaper cost than Reasoning models. Since most reasoning models can and probably do outperform decision models already in accuracy but they are a lot slower and are pricier.

Unless it can handle some different form of input other than just image and text, I don't see the point of having both flagship and flash version for the same model.

Re:Perplexity Decider v1.3: Training, eval, and use cases by declanjackson on Friday November 13, 2026 @06:01PM (#50491378)
But isn't there still a meaningful spread of difficulty, risk tolerance, etc. in decision model relevant tasks? Especially at high volumes I think there would still be value in being able to make intelligence/cost/speed/calibration tradeoffs between models
Re:Perplexity Decider v1.3: Training, eval, and use cases by deepsquirrelnet on Friday November 13, 2026 @06:14PM (#50491518)
> So your flagship model must be fastest and most accurate.

That's rarely the way anything in life works though. Do you prioritize fastest OR most accurate? That's a question that depends on context, task difficulty, etc.

You can use a 17M parameter NLI model for simple stuff and get orders of magnitude faster decisions. But the intelligence of that model will be quite a bit weaker.

In some tasks you might be talking about very small differences in accuracy between that model and a 27B parameter LLM decision model for several orders of magnitude in speed. In other tasks, the 17M might be about as good as a random guess.

Re:Perplexity Decider v1.3: Training, eval, and use cases by k__ on Friday November 13, 2026 @03:32PM (#50489235)
Does anyone really use such things for security?
Re:Perplexity Decider v1.3: Training, eval, and use cases by waximabbax on Friday November 13, 2026 @03:41PM (#50489357)
I highly doubt it. Even if a company does internally, its highly unlikely they will reveal it publically.

One use case is live content moderation or streaming where reasoning models could be a lot slower and its hard to hardcode logic.

Re:Perplexity Decider v1.3: Training, eval, and use cases by quinndupont on Friday November 13, 2026 @07:34PM (#50492179)
No, but great for text mining
Re:Perplexity Decider v1.3: Training, eval, and use cases by thatwasunusual on Saturday November 14, 2026 @03:14AM (#50495117)
How so? Honest question, because I can't think of a way to use a decision model for text mining...?
Re:Perplexity Decider v1.3: Training, eval, and use cases by 0o777 on Friday November 13, 2026 @03:41PM (#50489355)
Re:Perplexity Decider v1.3: Training, eval, and use cases by jstummbillig on Friday November 13, 2026 @04:00PM (#50489579)
Any ideas what decider models are actually good? There does not seem to be a strong competitive benchmark culture (as with LLMs) for whatever reason?
Re:Perplexity Decider v1.3: Training, eval, and use cases by walls on Friday November 13, 2026 @04:02PM (#50489599)
I've found this fairly accurate: https://huggingface.co/spaces/multimodalart/jev-decision-ind...

OneJev has been pretty amazing to run locally.

Re:Perplexity Decider v1.3: Training, eval, and use cases by viccis on Friday November 13, 2026 @05:23PM (#50490879)
That's good but I would definitely appreciate one with cost to run the benchmark and cost vs accuracy breakdowns.
Re:Perplexity Decider v1.3: Training, eval, and use cases by declanjackson on Friday November 13, 2026 @05:49PM (#50491233)
Artificial Analysis has an Index which also includes Cost per Task, Time per Task, accuracy, calibration etc. here: https://artificialanalysis.ai/evaluations/artificial-analysi...

(caveat, I work at AA and helped build this Index)

Re:Perplexity Decider v1.3: Training, eval, and use cases by talloaktrees on Friday November 13, 2026 @06:07PM (#50491426)
just found this today and used it as one of the links I sent a coworker who hadn't heard of decision models yet!
Re:Perplexity Decider v1.3: Training, eval, and use cases by stri8ted on Friday November 13, 2026 @04:01PM (#50489593)
It performs well on my local benchmark, and is priced competitively. Good timing too, considering that inception Labs is discontinuing their Mercury models API, which I was previously using. It seems plausible that the bidirectional attention performs better for decision models.
Re:Perplexity Decider v1.3: Training, eval, and use cases by nchmy on Friday November 13, 2026 @08:10PM (#50492460)
Because of this post, i discovered and dug into various decider and other models. The "locate" mechanism in Decider 1.3 seems to have an equivalent in Clairic from cloudflare and Extract from ScaleDown. I'm particularly impressed with scaledown for my use case - much cheaper and more accurate than the others.

https://scaledown.ai

https://blog.cloudflare.com/clairic-extraction-model/

https://blog.cloudflare.com/clef-decision-models/

Re:Perplexity Decider v1.3: Training, eval, and use cases by senjin on Friday November 13, 2026 @08:54PM (#50492780)
Perplexity Decider is my favorite of these new decision models but this version is a step back in my benchmark for my use case. The decision index benchmark[1] actually agrees with this and makes me think I should look at a higher ranked model

https://huggingface.co/spaces/multimodalart/jev-decision-ind...

Re:Perplexity Decider v1.3: Training, eval, and use cases by borg16 on Saturday November 14, 2026 @08:06AM (#50496726)
I am curious about your use case here, can you elaborate?
Re:Perplexity Decider v1.3: Training, eval, and use cases by neronuser on Saturday November 14, 2026 @04:44AM (#50495566)
Waiting for decision models being available from services like Bedrock. AWS recently published a post where they used decision models to boost evaluation performance, but they had to self-host it or call Jev through a Lambda
Re:Perplexity Decider v1.3: Training, eval, and use cases by throwaway_5753 on Saturday November 14, 2026 @07:41AM (#50496578)
V1.3 shows gains over v1.2 and Jev 1.14 on our internal evals for an entity extraction/matching task. Worth scoring it for your use case!
Re:Perplexity Decider v1.3: Training, eval, and use cases by Oras on Saturday November 14, 2026 @09:09AM (#50497157)
It’s fine considering the cost. I’ve run tests to classify lines from job ads to see which category they belong to, and have to say most of these models are disappointing. I suppose they were trained on long text not short ones.

Based on my tests, the top two (based on quality of classification and confidence level) were OpenAI decision and Liquid D1