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

Codestral Mamba

mistral.ai

Post reply on HN
Print view View original

51–60 of 148 posts

Re: Codestral Mamba

#51

Post by refulgentis »

passion__desire wrote: post #48
refulgentis wrote: post #29

Earlier quoted context omitted.

"All you need is users" doesn't seem optimal IMHO, Stability.ai providing an object lesson in that. They just released weights, and being a for profit, need to optimize for making money, not eyeballs. It seems wise to guide people to the API offering.

But they also signal competence in the space which means M&A. Or big nation states in future would hire them to produce country models once the space matures as was Emad's vision.

Did Emad's vision end up manifest? ex. did a nation-state end up paying Stability for a country model?

Would it help signal competency? They're a small team focused on making models, not VS Code extensions.

Would they do M&A? The founding team is ex-Googlers and has found significant attention in the MBA world via being an EU champion.

Re: Codestral Mamba

#52

Post by refulgentis »

bhouston wrote: post #46
refulgentis wrote: post #29

Earlier quoted context omitted.

"All you need is users" doesn't seem optimal IMHO, Stability.ai providing an object lesson in that. They just released weights, and being a for profit, need to optimize for making money, not eyeballs. It seems wise to guide people to the API offering.

On top of Hacker News (the target demographic for coders) without an effective monetizable call to action? What a missed opportunity. Github Copilot makes +100M/year, if not way way more. Having a VS Code extension for Mistral would be a revenue stream if it was one-click and better or cheaper than Github Copilot. It is malpractice in my mind to not be doing this if you are investing in creating coding models.

I see, that makes sense: make an extension and charge for it.

I assumed they meant free x local. It doesn't seem rational to make this one paid: its significantly smaller than their better model, and even more so than Copilot's.

Re: Codestral Mamba

#53

Post by red2awn »

monkeydust wrote: post #10

Any recommended product primers to Mamba vs Transformers - pros/cons etc?

A very good primer to state-space models (from which Mamba is based on) is The Annotated S4 [1]. If you want to dive into the code I wrote a minimal single-file implementation of Mamba-2 here [2].

[1]: https://srush.github.io/annotated-s4/

[2]: https://github.com/tommyip/mamba2-minimal

Re: Codestral Mamba

#54

Post by mbowcut2 »

causal wrote: post #38
smith7018 wrote: post #26

Earlier quoted context omitted.

Which means Codestral Mamba and DeepSeek both lead four benchmarks. Kinda takes the air out the announcement a bit.

It should be corrected but the interesting aspect of this release is the architecture. To stay competitive while only needing linear inference time and supporting 256k context is pretty neat.

THIS. People don't realize the importance of Mamba competing on par with transformers.

Re: Codestral Mamba

#55

Post by pzo »

weird they compare to deepseek-coder v1.5 when we already have v2.0. Any advantage to use codestral mamba apart from that it's lighter in weights?

Re: Codestral Mamba

#56

Post by mountainriver »

rjurney wrote: post #50
ukuina wrote: post #36

Earlier quoted context omitted.

What made you switch?

Claude is much better. Overwhelmingly better. It not only implements deep learning models for me, it has great suggestions on evolving them to actually work.

lol no it’s not, the benchmarks don’t show that at all. Both have issues in different ways

Re: Codestral Mamba

#58

Post by throwup238 »

ukuina wrote: post #36
rjurney wrote: post #16

But I JUST switched from GPT4o to Claude! :( Kidding, but it isn't clear how to use this thing, as others have pointed out.

What made you switch?

Claude Projects which allow attaching a bunch of files to fill up the 200k context. I wrote up a script to dump a bunch of code and documentation files to markdown as context and I add them to a bunch of Claude projects on a per topic basis.

For example, I'm currently working on a Rust/Qt desktop app so I have a project with the whole Qt6 book attached to ask questions about Qt, a project with my SQL schema and ORM/Sqlite docs to ask questions about the app's data and generate models without dealing with hallucinations, a project with all my QML files and Rust QML element code, a project with a bunch of Rust crate docs, and so on and on.

GPTs allow attaching files too but Claude Projects dump the entire contents of the files into the context rather than trying to do some hacky RAG that never works like I want it to.

Re: Codestral Mamba

#60

Post by causal »

mountainriver wrote: post #56
rjurney wrote: post #50

Earlier quoted context omitted.

Claude is much better. Overwhelmingly better. It not only implements deep learning models for me, it has great suggestions on evolving them to actually work.

lol no it’s not, the benchmarks don’t show that at all. Both have issues in different ways

Benchmarks are pretty flawed IMO, in particular their weakness here seems to be that they are poor at evaluating long-tail multiturn conversations. 4o often gives a great first response, then spirals into a repetition. Sonnet 3.5 is much better at seeing the big picture in a longer conversation IMO.
Post reply on HN