#8b
LightNews LightNews
chrisdeleon.bsky.social
curious if brave browser search was using openAI or claude to tell who I'm mad at. found announcement on how theirs won't make stuff up 🙃

that was a 2023 post but 2026 reddit suggest it's their own model (summarizer isn't Leo AI, second killswitch). even open source private LLMs are dangerous trash
Brave Scarch introduces the Summarizer, an AI tool for synthesized, relevant results PUBLISHED MAR 2, 2023 in Brave Search news Today we're thrilled to announce the latest Al-powered feature in Brave Search: the Summarizer. The Summarizer provides concise and to-the-point answers at the top of Brave Search results pages, in response to the user's input, solely based on Web search results. Unlike a purely generative Al model, which is prone to spout unsubstantiated assertions, we trained our large language models (LLMs) to process multiple sources of information present on the Web. This produces a more concise, accurate answer, expressed in coherent language. reddit post: Brave uses a variety of large language models (LLMs) across its Al products, with different models available depending on the service and user tier. For Brave Leo, the Al assistant, users can choose from models like Qwen 14B, Llama 3.1 8B, Claude 3.5 Haiku, Claude Sonnet 4, DeepSeek R1, and Gemma 12B, with the free version offering more limited access and the Premium version providing higher rate limits and access to larger models. The default model for Leo is Mixtral 8x7B, which is used for improved performance and speed. For Brave Search's Answer with Al, the system uses a combination of Llama and Qwen models, both open-source and hosted in-house to ensure privacy. Additionally, Brave has introduced Al Browsing mode in Brave Nightly, which leverages agentic Al to perform tasks like summarizing pages and navigating the web, with models selected based on the task and user settings.
September 29, 2026 at 6:37 PM
bireeves95.bsky.social
A Great idea……

We spend $8B a year hitting wildlife. If we used just a fraction of that to build 200 wildlife crossings annually, roadkill could be drastically reduced, making roads safer for animals and drivers alike.
September 29, 2026 at 6:29 PM
otoyori-blog.bsky.social
[BOT] 【音の寄り道】ブログ記事
音楽理論その4-1「ダイアトニックコード」
https://sounds-detour.com/2018/11/12/%e9%9f%b3%e6%a5%bd%e7%90%86%e8%ab%96%e3%81%9d%e3%81%ae4-1%e3%83%80%e3%82%a4%e3%82%a2%e3%83%88%e3%83%8b%e3%83%83%e3%82%af%e3%82%b3%e3%83%bc%e3%83%89/
September 29, 2026 at 6:03 PM
asamnews.bsky.social
3 Asian American non-profits back $8B Queens casino – AsAmNews
Several AAPI organizations have signed agreements with the Met Park casino plan, citing economic benefits while critics warn of addiction and rising rents.
asamnews.com
September 29, 2026 at 5:27 PM
lacus145.bsky.social
16th Anniversary Wardog Squadron Assisted The 3rd Osean Naval Fleet in Eaglin Straits and Soon The Yuktobanian Navy and Scinfaxi Sent F-35C AV-8B. The Battle Soon Raged, SLBM Missiles Sank 2 Aircraft Carriers.

Rendezvous Operation Whalebird
30 September 2010
September 29, 2026 at 5:26 PM
wochitube.bsky.social
「ヤニねこ」←このアニメが人気跳ねなかった理由www

https://takaiotaku.com/%e3%80%8c%e3%83%a4%e3%83%8b%e3%81%ad%e3%81%93%e3%80%8d%e2%86%90%e3%81%93%e3%81%ae%e3%82%a2%e3%83%8b%e3%83%a1%e3%81%8c%e4%ba%ba%e6%b0%97%e8%b7%b3%e3%81%ad%e3%81%aa%e3%81%8b%e3%81%a3%e3%81%9f%e7%90%86/
「ヤニねこ」←このアニメが人気跳ねなかった理由www
何故なのか
takaiotaku.com
September 29, 2026 at 5:21 PM
jeremykun.com
Oops, I mis-remembered; it's 8s/token for Bert-base, 130s/token for Llama-8B. Still impressive!
September 29, 2026 at 4:34 PM
huggingfacetrends.bsky.social
今日のHuggingFaceトレンド

Contrastive-LM/CLM-v0.1-8B
Qwen3-8Bをベースにした対照学習モデルCLMを提供することを目的としています。
状態と行動を結びつけるSystem Oneモデルとして設計されており、低遅延かつ高速なランキングや分類を実現します。
エージェントの軌跡データを用いて学習されており、従来のモデルより高速な動作が可能です。
Contrastive-LM/CLM-v0.1-8B · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
September 29, 2026 at 4:21 PM
jeremykun.com
Congrats to my colleagues at CMU for releasing their Cerium library: github.com/CMU-CAOS/Cer...

It's a multi-GPU FHE (CKKS) implementation that achieved really impressive performance on Llama-8B inference (to the tune of 20 seconds per token). I am excited to integrate it into HEIR!
GitHub - CMU-CAOS/Cerium
Contribute to CMU-CAOS/Cerium development by creating an account on GitHub.
github.com
September 29, 2026 at 4:20 PM
ruidomag.bsky.social
Swedish House Mafia 新單曲〈...

https://ruidomag.com/swedish-house-mafia-%e6%96%b0%e5%96%ae%e6%9b%b2%e3%80%88happiness-is-so-sad%e3%80%89mv%ef%bc%8cspotify-%e7%8d%a8%e5%ae%b6-48-%e5%b0%8f%e6%99%82%e6%90%b6%e5%85%88%e6%94%b6%e7%9c%8b/?utm_source=bluesky&utm_medium=social&utm_campaign=news
Swedish House Mafia 新單曲〈Happiness Is So Sad〉MV，Spotify 獨家 48 小時搶先收看
Swedish House Mafia 公開新單曲...
ruidomag.com
September 29, 2026 at 3:52 PM
lewis57.bsky.social
substack.com/@edwardelson...
Ed Elson (@edwardelson)
Anthropic's 2025 financials were just revealed: $6B revenue $8B operating loss $42B net loss 💀
substack.com
September 29, 2026 at 3:30 PM
mikabot.precure.ml.ap.brid.gy
【予定】1日前
10/01 19:30〜20:00
Go！プリンセスプリキュア 47話 花のように…！つよくやさしく美しく！

TOKYO MX再放送
https://s.mxtv.jp/anime/princess_precure/

タグ付けの設定手順
キュアスタ！ https://blog.precure.ml/articles/%E5%AE%9F%E6%B3%81%E7%94%A8%E3%82%BF%E3%82%B0%E3%82%92%E8%A8%AD%E5%AE%9A%E3%81%99%E3%82%8B/
きゅあすきー […]
Original post on precure.ml
precure.ml
September 29, 2026 at 3:17 PM
mikabot.precure.ml.ap.brid.gy
❗️本日❗️
09/30 19:30〜20:00
わんだふるぷりきゅあ！ 49話 あなたの声

TOKYO MX再放送
https://s.mxtv.jp/anime/wonderful_precure/

タグ付けの設定手順
キュアスタ！ https://blog.precure.ml/articles/%E5%AE%9F%E6%B3%81%E7%94%A8%E3%82%BF%E3%82%B0%E3%82%92%E8%A8%AD%E5%AE%9A%E3%81%99%E3%82%8B/
きゅあすきー […]
Original post on precure.ml
precure.ml
September 29, 2026 at 3:17 PM
sichgate.bsky.social
We tested 7 open-weight models at fp16 and 4-bit NF4. Same 179 probes.

After quantization:
5 got worse (Mistral-7B +7.26 pts)
1 didn't move (Phi-4-mini)
1 got better (Granite 3.0 8B, -4.47)

Same technique, three outcomes. Test the version you ship.
September 29, 2026 at 3:01 PM
yurinavi.bsky.social
Jp9LGb-G3uBg5O_mM&dib_tag=se&keywords=%E3%80%8C%E3%81%88%E3%80%81%E7%A7%81%E3%81%8C%E3%81%9D%E3%81%A3%E3%81%A1%E3%81%AA%E3%81%AE%21%3F%E3%80%8D%E5%AD%90%E7%8A%AC%E7%B3%BB%E5%BE%8C%E8%BC%A9%E5%BD%BC%E5%A5%B3%E3%81%AB%E9%A3%9F%E3%81%B9%E3%82%89%E3%82%8C%E3%82%8B%E3%82%AB%E3%83%83%E3%82%B3%E3%81%84%E3%
September 29, 2026 at 2:31 PM
yurinavi.bsky.social
▼amazonページ
https://www.amazon.co.jp/%E3%80%8C%E3%81%88%E3%80%81%E7%A7%81%E3%81%8C%E3%81%9D%E3%81%A3%E3%81%A1%E3%81%AA%E3%81%AE-%E3%80%8D%E5%AD%90%E7%8A%AC%E7%B3%BB%E5%BE%8C%E8%BC%A9%E5%BD%BC%E5%A5%B3%E3%81%AB%E9%A3%9F%E3%81%B9%E3%82%89%E3%82%8C%E3%82%8B%E3%82%AB%E3%83%83%E3%82%B3%E3%81%84%E3%81%84%E
September 29, 2026 at 2:31 PM
Amazon is getting a piece of Generac for buying $8B worth of gens for Amazon data center backup.
September 29, 2026 at 2:26 PM
hayase.science
公開されたばかりのllm-jp-4.1ですが、公式ggufはGUIアプリケーションでそのまま動作せず使い勝手が悪いため、ひとまず低VRAMでも高速で動く8b 4bit量子化モデルを作りました。unsloth desktopに直接導入可能です。
huggingface.co/genh215/llm-...
genh215/llm-jp-4.1-8b-thinking-bnb-4bit · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
September 29, 2026 at 2:07 PM
bdunwood.bsky.social
DeepSeek's V4.1 Flash Beats Its Own Flagship Model, Costs 20x Less Than Claude Opus 5
DeepSeek launched V4.1 Flash with a new architecture that uses 8B active parameters for input and 16B for output. V4-Pro is being phased out September 14. Continue reading...
www.vktr.com
September 29, 2026 at 2:00 PM
vktrnow.bsky.social
DeepSeek's V4.1 Flash Beats Its Own Flagship Model, Costs 20x Less Than Claude Opus 5: DeepSeek launched V4.1 Flash with a new architecture that uses 8B active parameters for input and 16B for output. V4-Pro is being phased out September 14.
Continue reading... #aiplatforms #deepseek
DeepSeek's V4.1 Flash Beats Its Own Flagship Model, Costs 20x Less Than Claude Opus 5
DeepSeek launched V4.1 Flash with a new architecture that uses 8B active parameters for input and 16B for output. V4-Pro is being phased out September 14. Continue reading...
www.vktr.com
September 29, 2026 at 2:00 PM
ai-katuyo-lab.bsky.social
AIの月額課金、増やしたくない人へ👀
1minAIなら…
✅大型モデルを無料枠でお試し
✅Qwen3-8Bは無料で使い放題
✅APIも無料クレジットで利用可能
✅エージェントで面倒な指示づくりを省力化

StackSocialには買い切りプランも！
月額契約を増やさず、AIを日常の相棒に✨
※買い切りはクレジット制・サービス提供期間中。

まずは無料で体験👇
1min.ai?referrer_id=...
#1minAI #生成AI #AI活用 #API #非サブスク #買い切り #無料 #Claude #Gemini　#ChatGPT #DeepSeek
1minAI | Free All-in-one AI App for Text, Image, Audio, Video
Unlock the power of AI for all your creative needs. Our free all-in-one platform empowers you with AI tools for text, writing, image, audio, and video.
1min.ai
September 29, 2026 at 1:47 PM
yurinavi.bsky.social
▼amazonページ
https://www.amazon.co.jp/%E9%80%B1%E3%81%AB%E4%B8%80%E5%BA%A6%E3%82%AF%E3%83%A9%E3%82%B9%E3%83%A1%E3%82%A4%E3%83%88%E3%82%92%E8%B2%B7%E3%81%86%E8%A9%B110-%E3%81%B5%E3%81%9F%E3%82%8A%E3%81%AE%E7%A7%98%E5%AF%86%E3%81%AF%E4%B8%80%E3%81%A4%E5%B1%8B%E6%A0%B9%E3%81%AE%E4%B8%8B-%E3%83%95%E3%82%A
September 29, 2026 at 1:31 PM
earanyash.bsky.social
AMD is buying World Labs, the spatial AI startup Fei-Fei Li started barely a year ago, for more than $8 billion in an all stock deal.

Nvidia has owned the AI chip story for years. AMD just bought its way into the next one: AI that understands 3D space.
AMD buys Fei-Fei Li's AI startup World Labs for $8B. AMD is buying World Labs, a startup founded by Fei-Fei Li that builds AI for understanding 3D space, in an all-stock deal worth more than $8 billion.
September 29, 2026 at 1:01 PM