Hacker News: nabakin

New comment by nabakin in "Jamesob's guide to running SOTA LLMs locally"

nabakin — Fri, 03 Jul 2026 19:52:31 +0000

Are you running qwen3.6-27b on one 3090 with your KV cache at q4? Ime there is significant long-context recall accuracy degradation at that precision. I prefer putting the KV cache at q8 and working with the 120k context

New comment by nabakin in "U.S. DOJ demands Apple and Google unmask over 100k users of car-tinkering app"

nabakin — Fri, 15 May 2026 21:29:52 +0000

Then you can sandbox

New comment by nabakin in "DeepSeek v4"

nabakin — Sun, 26 Apr 2026 12:39:49 +0000

I think they were mistaken or maybe they were just referring to inference because I don't see anyone making that claim and it would be quite the news.

New comment by nabakin in "DeepSeek v4"

nabakin — Sat, 25 Apr 2026 00:38:58 +0000

Probably because you said you used DeepSeek. People don't want to see AI in the comments and don't trust AI responses.

New comment by nabakin in "DeepSeek v4"

nabakin — Fri, 24 Apr 2026 18:01:29 +0000

Yes, that's the footnote from citation [5].

New comment by nabakin in "DeepSeek v4"

nabakin — Fri, 24 Apr 2026 16:55:26 +0000

> Also, note that there's zero CUDA dependency. It runs entirely on Huawei chips.

That is a huge claim to make with no evidence.

I researched what you said, and I have found no statement to that effect in their paper[0], on huggingface[1], twitter[2], WeChat[3], or in their news release[4].

They only mention as a footnote in only the Chinese version of their news release that they plan to reduce inference costs with the Ascend 950 supernode when it releases[5]. The only mention of Huawei in their paper is that they validated a technique to lower interconnect bandwidth on Ascend NPUs and Nvidia GPUs[6].

[0] https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main...

[1] https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro

[2] https://xcancel.com/deepseek_ai/status/2047516922263285776

[3] https://mp.weixin.qq.com/s/8bxXqS2R8Fx5-1TLDBiEDg

[4] https://api-docs.deepseek.com/news/news260424

[5] https://api-docs.deepseek.com/zh-cn/img/v4-price.png

[6] Page 16

New comment by nabakin in "NIST scientists create 'any wavelength' lasers"

nabakin — Sun, 19 Apr 2026 00:14:09 +0000

> When it comes to information transfer and processing, light can do things that electricity can’t. Photons — particles of light — are far zippier than electrons at working their way through circuits.

Electrons themselves don't move at the speed of light, but information transfer (i.e. communication) via electrons does happen close to the speed of light.

A subtle, but important, distinction that's often misunderstood and means computational performance gains would probably come from bandwidth, not latency.

New comment by nabakin in "Google releases Gemma 4 open models"

nabakin — Thu, 02 Apr 2026 20:55:05 +0000

Dunno but there's a PR for it. Probably also more performant than Modular.

New comment by nabakin in "Google releases Gemma 4 open models"

nabakin — Thu, 02 Apr 2026 19:34:09 +0000

If OP meant they have the fastest implementation of Gemma 4 on Blackwell at the moment, I guess that is technically true. I doubt that will hold up when TensorRT-LLM finishes their implementation though.

New comment by nabakin in "Google releases Gemma 4 open models"

nabakin — Thu, 02 Apr 2026 19:30:14 +0000

I know Arc AGI 2 has a private test set and they have a good amount of results[0] but it's not a conventional benchmark.

Looking around, SWE Rebench seems to have decent protection against training data leaks[1]. Kagi has one that is fully private[2]. One on HuggingFace that claims to be fully private[3]. SimpleBench[4]. HLE has a private test set apparently[5]. LiveBench[6]. Scale has some private benchmarks but not a lot of models tested[7]. vals.ai[8]. FrontierMath[9]. Terminal Bench Pro[10]. AA-Omniscience[11].

So I guess we do have some decent private benchmarks out there.

[0] https://arcprize.org/leaderboard

[1] https://swe-rebench.com/about

[2] https://help.kagi.com/kagi/ai/llm-benchmark.html

[3] https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard

[4] https://simple-bench.com/

[5] https://agi.safe.ai/

[6] https://livebench.ai/

[7] https://labs.scale.com/leaderboard

[8] https://www.vals.ai/about

[9] https://epoch.ai/frontiermath/

[10] https://github.com/alibaba/terminal-bench-pro

[11] https://artificialanalysis.ai/articles/aa-omniscience-knowle...

New comment by nabakin in "Google releases Gemma 4 open models"

nabakin — Thu, 02 Apr 2026 18:00:00 +0000

It's easy to game and human evaluation data has its trade-offs, but it's way easier to fake public benchmark results. I wish we had a source of high quality private benchmark results across a vast number of models like Lmarena. Having high quality human evaluation data would be a plus too.

New comment by nabakin in "Google releases Gemma 4 open models"

nabakin — Thu, 02 Apr 2026 17:43:46 +0000

Faster than TensorRT-LLM on Blackwell? Or do you not consider TensorRT-LLM open source because some dependencies are closed source?

New comment by nabakin in "Google releases Gemma 4 open models"

nabakin — Thu, 02 Apr 2026 17:17:26 +0000

It's referring to the Lmsys Leaderboard/Lmarena/Arena.ai[0]. It's very well-known in the LLM community for being one of the few sources of human evaluation data.

[0] https://arena.ai/leaderboard/chat

New comment by nabakin in "Google releases Gemma 4 open models"

nabakin — Thu, 02 Apr 2026 17:00:46 +0000

Public benchmarks can be trivially faked. Lmarena is a bit harder to fake and is human-evaluated.

I agree it's misleading for them to hyper-focus on one metric, but public benchmarks are far from the only thing that matters. I place more weight on Lmarena scores and private benchmarks.

New comment by nabakin in "Big data on the cheapest MacBook"

nabakin — Thu, 12 Mar 2026 16:16:10 +0000

And nowadays we have Debian running in a VM on Android [1]

[1] https://www.zdnet.com/article/how-to-use-the-new-linux-termi...

New comment by nabakin in "MacBook Pro with M5 Pro and M5 Max"

nabakin — Tue, 03 Mar 2026 16:49:08 +0000

I would consider it reasonable if this was 4x TTFT and Throughput, but it seems like it's only for TTFT.

New comment by nabakin in "Tesla kills Autopilot, locks lane-keeping behind $99/month fee"

nabakin — Fri, 23 Jan 2026 20:11:20 +0000

And right to repair

New comment by nabakin in "JetBlue flight averts mid-air collision with US Air Force jet"

nabakin — Tue, 16 Dec 2025 18:02:02 +0000

Ty this is great

New comment by nabakin in "JetBlue flight averts mid-air collision with US Air Force jet"

nabakin — Tue, 16 Dec 2025 02:39:36 +0000

TIL Europe still has some presence in the Americas. Thought all of that was gone with the Monroe Doctrine

New comment by nabakin in "France threatens GrapheneOS with arrests / server seizure for refusing backdoors"

nabakin — Mon, 24 Nov 2025 22:11:23 +0000

Fyi it doesn't look like this post is listed on the frontpage anymore, even with the points it has. Not sure if it's intentional