Hackernews posts about Llama 4

RSS feed for Llama 4

01
newShow HN: Llama.cpp fork with 2-4x multiGPU speed for MoE models bigger than VRAMgithub.com
▲ 2neuralll 10d 💬 3
02
newGetting 50 GB/S Back from the Apple Neural Engineeiln.github.io
▲ 218eiln 26d 💬 36
03
newLaunch HN: Magnitude (YC S25) – Self-optimizing inference engine for agentsgithub.com
▲ 194anerli 5d 💬 99
04
newWhy human syntax breaks LLMs (and how to fix agentic coding)
▲ 5aslang 28d 💬 3
05
newShow HN: IttyBittyAI – Qwen2.5:0.5B running on a 2017 Asus smartphone
▲ 2spottedmarley 11d discuss
06
newShow HN: PicoLM v1.0-rc2 ("Yura Kana"). Run an LLM on Digital Unixgithub.com
▲ 1gabucino 12d discuss
07
newOllama replacement 2-4x faster for no extra compute costgithub.com
▲ 4wowitsbase 28d 💬 4
08
newShow HN: Breadcrumb, record everything on your mac + context manager for AIinnerloop.works
▲ 49jv22222 4d 💬 9
09
newShow HN: Ollama stand in replacement, 2-4x faster, not just basic optimizationsgithub.com
▲ 3wowitsbase 27d discuss
10
newShodan indexes over 47.000 exposed Ollama instancestwitter.com
▲ 2soltanov 19d 💬 1
11
newHi HN
▲ 3DEVGUSTAVOSIS 17d 💬 1
12
newThe Llama 4 herdai.meta.com
▲ 1235georgehill 548d 💬 658
13
newLlama 4 Now Live on Groqgroq.com
▲ 109gok 548d 💬 48
14
newMeta AI App built with Llama 4about.fb.com
▲ 98friggeri 524d 💬 109
15
newCerebras achieves 2,500T/s on Llama 4 Maverick (400B)www.cerebras.ai
▲ 93ByteAtATime 493d 💬 93
16
newAsk HN: 10 months since the Llama-4 release: what happened to Meta AI?
▲ 55Invictus0 242d 💬 12
17
newThe Llama 4 Herdai.meta.com
▲ 48ushakov 548d 💬 1
18
newLlama 4 Smells Badfastml.com
▲ 41alexmolas 530d 💬 26
19
newYann LeCun confirms Meta's Llama 4 benchmarks were "fudged a little bit"tech.slashdot.org
▲ 30frays 274d 💬 2
20
newLlama 405B 506 tokens/second on an H200developer.nvidia.com
▲ 21moondistance 722d 💬 5
21
newLlama 4 Is Banned in the EU: Open AI, Region-Lockeddionwiggins.substack.com
▲ 10simjue 540d 💬 6
22
newMeta's surprise Llama 4 drop exposes the gap between AI ambition and realityarstechnica.com
▲ 9pera 546d discuss
23
newMeta is using more than 100k Nvidia H100 AI GPUs to train Llama-4www.tomshardware.com
▲ 8turtlegrids 703d 💬 1
24
newSerious issues in Llama 4 training. VP of AI at Meta resignedold.reddit.com
▲ 7devops000 546d 💬 3
25
newBias Reduced by 44% in Llama 4 Using Machine Unlearningwww.hirundo.io
▲ 6benlur 543d discuss
26
newLlama 4: Did Meta just push the panic button?www.interconnects.ai
▲ 6lairv 546d discuss
27
newLlama 4 performs worse than Llama 3 at translationnuenki.app
▲ 5Alex-Programs 548d 💬 3
28
newLlama 4 underperforms: a benchmark against coding-centric modelsrootly.com
▲ 5sylvainkalache 539d discuss
29
newMeta exec denies the company artificially boosted Llama 4's benchmark scorestechcrunch.com
▲ 5pera 546d discuss
30
newReleased Llama 4 Maverick places 32nd in LMArenalmarena.ai
▲ 4joaogui1 543d 💬 1
31
newLlama 4: One Week Afterblog.kilocode.ai
▲ 4ofou 542d discuss
32
newInitial Impressions of Llama 4simonwillison.net
▲ 4ode 548d discuss
33
newShow HN: Fast and Cheap Llama-405Bcentml.ai
▲ 4akbarnur 700d discuss
34
newShow HN: How to guide on training Llama-405B using PyTorch distributed APIsgithub.com
▲ 3lambda-research 720d 💬 4
35
newLimited-time GPU firepower Dirt-cheap LLM Inference: Llama 4, DeepSeek 0528www.cloudrift.ai
▲ 3dikobraz 484d 💬 1
36
newFacebook Pushes Its Llama 4 AI Model to the Right, Wants to Present "Both Sides"www.404media.co
▲ 3easyThrowaway 541d 💬 1
37
newLlama 4 available free and unlimitedwww.blackbox.ai
▲ 3mcflem007 548d 💬 1
38
newUnmodified Llama 4 Maverick ranks below rivals after Meta cheating allegationswww.neowin.net
▲ 3bundie 542d discuss
39
newLlama 4 performance on Kotlin/Android coding tasksfirebender.com
▲ 3kevo1ution 545d discuss
40
newMeta cheats on Llama 4 benchmarkwww.heise.de
▲ 3Topfi 545d discuss
41
newMeta Releases Llama 4: Here's the Hardware You'll Need to Run It Yourselfwww.hardware-corner.net
▲ 3mdp2021 547d discuss