Hackernews posts about 405B

1. 01
   

   newShow HN: Designobo – AI prototyping tool in 40kb of jsdesignobo.com

   ▲ 1crowdyriver11ddiscusssavesaved
2. 02
   

   newSHOW HN: I built the fastest PHP webserver in the world

   ▲ 6EGreg29d💬 1savesaved
3. 03
   

   newBig Tech uses guarantees to keep $300B AI exposure off balance sheetswww.ft.com

   ▲ 38sbulaev26d💬 67savesaved
4. 04
   

   newEtched fields funding offers at $40B+ valuationtechcrunch.com

   ▲ 12gmays10d💬 4savesaved
5. 05
   

   newWhat is the AI capex breakeven rate?www.ft.com

   ▲ 3AnodicElegy18d💬 2savesaved
6. 06
   

   newShow HN: Open Twin, a modelica-based simulator for real-world physical systemsopen-twin.netlify.app

   ▲ 1tgirotto11ddiscusssavesaved
7. 07
   

   newUK's Extend Robotics Secures US$3.3M to Sell Factory Work Instead of Machineswww.saasrise.com

   ▲ 1monkeydust11ddiscusssavesaved
8. 08
   

   newIs Ken Griffin turning Wynwood into a college town?www.miaminewtimes.com

   ▲ 1wslh16ddiscusssavesaved
9. 09
   

   newWhat's going on with AI token price deflation?www.ft.com

   ▲ 1AnodicElegy11d💬 1savesaved
10. 10
    

    newLlama 3.1 405B now runs at 969 tokens/s on Cerebras Inferencecerebras.ai

    ▲ 427benchmarkist696d💬 156savesaved
11. 11
    

    newHuggingFace - Tencent launches Hunyuan Large which outperforms Llama 3.1 405Bhuggingface.co

    ▲ 23janik-io710d💬 1savesaved
12. 12
    

    newAllenAI Tulu 3 405B available for chat and download

    ▲ 12soundworlds619ddiscusssavesaved
13. 13
    

    newTülu 3 405B – GPT-4o performance with open-source data, code, and recipestwitter.com

    ▲ 6rvnx624ddiscusssavesaved
14. 14
    

    newRunning Llama 3.1 405Ba11ce.com

    ▲ 4a11ce49ddiscusssavesaved
15. 15
    

    newShow HN: Fast and Cheap Llama-405Bcentml.ai

    ▲ 4akbarnur711ddiscusssavesaved
16. 16
    

    newNous Research – Hermes 4 405B/70B releasedhermes4.nousresearch.com

    ▲ 3beklein416d💬 1savesaved
17. 17
    

    newShisa V2 405B: Japan's Highest Performing LLMshisa.ai

    ▲ 3JimDabell500ddiscusssavesaved
18. 18
    

    newLlama 405B up to 142 tok/s on Nvidia H200 SXMold.reddit.com

    ▲ 2avianion718d💬 1savesaved
19. 19
    

    newHow to Run Meta Llama 3.1 405B with Nebius AI Studio APInebius.com

    ▲ 1dsaed718ddiscusssavesaved
20. 20
    

    newShow HN: Open-Source Python REPL with AI Tutor for Learning and Problem-Solvingwww.companionai.dev

    ▲ 5lambda-research716d💬 2savesaved
21. 21
    

    newAsk HN: AI Cloud Computing, is it cheaper than the OpenAI API in the end?

    ▲ 2calipsow722d💬 4savesaved
22. 22
    

    newAsk HN: How long before we get "coding agent in a box"?

    ▲ 2heavyarms298d💬 3savesaved
23. 23
    

    newVakgpt: Open-Source Chat Wrapper for Rapid AI Prototyping

    ▲ 1krishna-vakx506ddiscusssavesaved
24. 24
    

    newShow HN: Most Efficient Batch API for Open-Source and Custom Modelswithexxa.com

    ▲ 1Blue_Cosma728ddiscusssavesaved
25. 25
    

    newScaling the Tülu 3 post-training recipes to surpass the perf of DeepSeek V3allenai.org

    ▲ 14Philpax624ddiscusssavesaved
26. 26
    

    newCompetition intensifies for Anthropic and OpenAI ahead of IPOswww.ft.com

    ▲ 2JumpCrisscross112ddiscusssavesaved
27. 27
    

    newiPhone 17 Pro Demonstrated Running a 400B LLMtwitter.com

    ▲ 713anemll207d💬 326savesaved
28. 28
    

    newShow HN: Z80-μLM, a 'Conversational AI' That Fits in 40KBgithub.com

    ▲ 514quesomaster9000291d💬 122savesaved
29. 29
    

    newWe hid backdoors in ~40MB binaries and asked AI + Ghidra to find themquesma.com

    ▲ 245jakozaur236d💬 98savesaved
30. 30
    

    newTrinity large: An open 400B sparse MoE modelwww.arcee.ai

    ▲ 231linolevan261d💬 82savesaved
31. 31
    

    newICE test train reaches speeds of up to 405.0 km/hwww.deutschebahn.com

    ▲ 113doener474d💬 148savesaved
32. 32
    

    newCerebras achieves 2,500T/s on Llama 4 Maverick (400B)www.cerebras.ai

    ▲ 93ByteAtATime503d💬 93savesaved
33. 33
    

    newWD Bets 8x Bandwidth Beats More Terabytes as 40TB UltraSMR Starts Shippingwww.storagereview.com

    ▲ 18rbanffy69d💬 4savesaved
34. 34
    

    newRAG a 40GB Outlook inbox – Long term Staff member leaving, keeping knowledgeold.reddit.com

    ▲ 13miles652d💬 3savesaved
35. 35
    

    newShow HN: Q.js – Smaller than React/Vue, yet more powerful (40KB gzipped)github.com

    ▲ 12EGreg411d💬 20savesaved
36. 36
    

    newThe three lines of CSS that saved me 40kb and might do the same for youblog.welcomehome.city

    ▲ 11speckx127d💬 2savesaved
37. 37
    

    newNex N2 Pro: Frontier agentic performance at 400Bhuggingface.co

    ▲ 7himata4113130ddiscusssavesaved
38. 38
    

    newFujifilm to begin shipping 40TB Ultrium tape cartridges in January 2026www.tomshardware.com

    ▲ 6rbanffy291d💬 1savesaved
39. 39
    

    newItalian startup working on a 400B language model (Italian)www.ilsole24ore.com

    ▲ 5theanonymousone114ddiscusssavesaved

· edit