moondistance

5 ms·

1.
by moondistance 3mo ago
Paper: https://huggingface.co/papers/2606.16140
2.
VibeThinker-3B achieves 80.2 on LCBv6(twitter.com)
8 points by moondistance 3mo ago|3 comments
3.
Qwen 3.5: small models with impressive performance(twitter.com)
6 points by moondistance 7mo ago|0 comments
4.
Language Models Are Injective and Hence Invertible(arxiv.org)
1 points by moondistance 11mo ago|2 comments
5.
by moondistance 1y ago
How does this compare with Exowatt? Assume fresnel lenses are more efficient.
6.
Autonomous Trash Can(twitter.com)
20 points by moondistance 1y ago|1 comments
7.
Grok 4 Benchmarks(twitter.com)
3 points by moondistance 1y ago|0 comments
8.
by moondistance 1y ago
Chris Crawford is also famous for the Dragon Speech :) https://youtu.be/CBrj4S24074?si=Ph12RpW8BKsh8-qS
9.
by moondistance 1y ago
Yes, big time, but there continues to be lots of progress. Most importantly, models are maturing, and this means less custom optimization is required.
10.
by moondistance 1y ago
(this is also why AMD popped 10% at open yesterday - this is a new development and talks from their 2025 "Advancing AI" event were published late last week + over the weekend)
11.
by moondistance 1y ago
Yes, for many applications. Meta, OpenAI, Crusoe, and xAI recently announced large purchases of MI300 chips for inference. MI400, which will be available next year, also looks to be at least on par with Nvidia's roadmap.
12.
by moondistance 1y ago
This is from 2020.
13.
by moondistance 1y ago
Thanks :)
14.
by moondistance 1y ago
Where do you find/buy land? How do you vet purchases? Can you point to a few websites, etc.? Thanks!
15.
Jensen Huang – Nvidia GTC 2025 Keynote(nvidia.com)
75 points by moondistance 2y ago|74 comments
16.
Exaone Deep 32B – beats DeepSeek 671B(lgresearch.ai)
3 points by moondistance 2y ago|1 comments
17.
DeepSeek-R1 at 3,872 tokens / second on a single Nvidia HGX H200(blogs.nvidia.com)
13 points by moondistance 2y ago|1 comments
18.
ByteDance Doubao-1.5-pro matches GPT 4o benchmarks at 50x cheaper(twitter.com)
4 points by moondistance 2y ago|0 comments
19.
by moondistance 2y ago
Happy New Year! I know so few of you but feel like you’re my best friends. Love to all.
20.
US-China Commission top recommendation: Manhattan project for race to AGI [pdf](uscc.gov)
4 points by moondistance 2y ago|0 comments
21.
by moondistance 2y ago
Microsoft is negotiating equity in OpenAI as part of the switch to a for-profit. Non-zero chance this is a negotiation flex.
22.
AI scans RNA 'dark matter' and uncovers 70k new viruses(nature.com)
1 points by moondistance 2y ago|0 comments
23.
Llama 405B 506 tokens/second on an H200(developer.nvidia.com)
21 points by moondistance 2y ago|5 comments
24.
by moondistance 2y ago
Significant further optimizations. FP8!
25.
by moondistance 2y ago
VP Research Barret Zoph and Chief Research Officer Bob McGrew also announced their departures this evening.
26.
SenseNova 5.5 claims SOTA LLM benchmark results(twitter.com)
2 points by moondistance 2y ago|0 comments
27.
New AI Training Technique Is Drastically Faster, Says Google(decrypt.co)
84 points by moondistance 2y ago|38 comments
28.
Nvidia open source LLM Nemotron 4 340B at top of the charts [pdf](d1qx31qr3h6wln.cloudfront.net)
17 points by moondistance 2y ago|1 comments
29.
Uncensor Any LLM with Abliteration(huggingface.co)
4 points by moondistance 2y ago|0 comments
30.
by moondistance 2y ago
I'd like to build a chatbot arena with pro-humanity questions.

More ›