frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Ask HN: Who wants to be hired? (September 2026)

100•whoishiring•1d ago•361 comments

Ask HN: Who is hiring? (September 2026)

234•whoishiring•1d ago•274 comments

Kimi K3 and GLM-5.3 are better than Gemini 3.8 Flash

7•theHocineSaad•3h ago•2 comments

ChatGPT Linux GUI version updates redownload all files, require 1.4GB download

7•juujian•8h ago•0 comments

Ask HN: How to ask questions in Ask HN

3•samikama•5h ago•8 comments

Ask HN: Would you be interested in using a VoIP system that has an operator?

2•left-struck•5h ago•4 comments

Ask HN: Where do you get your AI (technology) news?

4•BtM909•8h ago•0 comments

Resources to Get Good at Soldering?

4•tosmatos•11h ago•1 comments

Knowledge Graph Engine for Codebases

2•Shourya_55•6h ago•8 comments

Noviz AI Agent for ERPNext – A Reasoning Layer for ERPNext

2•tajdink•8h ago•0 comments

Whose interview process has caught up to the new way of working?

2•verygoodenginee•10h ago•5 comments

Ask HN: Anyone homeschooling their kids?

12•wbnns•20h ago•18 comments

Ask HN: What would happen if your company stopped using all AI tomorrow?

38•jc_811•2d ago•83 comments

Were Claude's Weekly Limits Cut in Half? (and obscured by another 50% "deal"?)

2•benwills•18h ago•3 comments

Dropbox Data Breach

46•hmate9•2d ago•7 comments

My local model setup on an M4 Pro Mac Mini

4•RanjithKumarRag•8h ago•1 comments

Ask HN: Does HN block mobile browsers?

4•pbohun•22h ago•7 comments

Is it just me, or has Claude Opus gotten worse recently?

12•de6u99er•1d ago•20 comments

Claude Code now appends a link to a Claude session in every commit

10•codexon•1d ago•2 comments

Why did Microsoft acquire LinkedIn?

6•ninju•1d ago•4 comments

Old.reddit.com no longer works for logged out users

31•kradeelav•1d ago•15 comments

Ask HN: Alternative to Proton Mail?

7•OhMeadhbh•1d ago•7 comments

Ask HN: What do your job interviews look like?

29•Hixon10•2d ago•10 comments

Asked Reddit support to give me API key so I can delete my data. They refused

15•leoriq•1d ago•18 comments

You've reached the end!

Open in hackernews

Kimi K3 and GLM-5.3 are better than Gemini 3.8 Flash

7•theHocineSaad•3h ago
Based on artificialanalysis.ai, Kimi K3 and GLM-5.3 are more intelligent than the new Gemini 3.8 Flash[1][2].

Gemini 3.8 Flash comes in eighth place with a score of 59, just after Kimi K3 and GLM-5.3, with scores of 60 for both of them.

Gemini 3.8 Flash was better in terms of Cost per task, with $0.58 spent to finish the task, $0.68 for GLM-5.3, and 0.84 for Kimi K3[3][4].

Because results may change over time, I included screenshots taken at the time of writing this post.

[1] https://artificialanalysis.ai/#intelligence [2] https://imgur.com/a/BMOJBED

[3] https://artificialanalysis.ai/#price-and-cost [4] https://imgur.com/a/55sQRUZ

Comments

smallerize•3h ago
You're comparing non-flash models to a flash model. They are more expensive, and K3 is a lot slower too.
verdverm•1h ago
qwen 3.8 27B is right there with them too (last week anyhow, had the #11 spot)