3,943 Photos and videos
State of Local AI #1
———
In lieu of Fable ban.
Here’s the best LLMs of the week to run on your hardware.
—— 4-8gb vram/ram 500$
- Gemma-4-qat huggingface.co/unsloth/gemma… I had someone mention it’s very good for subagent stuff
—— 8-16gb vram/ram < 1k usd
- Gemma-12B huggingface.co/google/gemma-… without a doubt the smartest model of its size
—— 16-32gb Apple/Strix halo 1-2k usd
- Diffusion Gemma26B huggingface.co/google/diffus…
- on 1x 6000 it’s eating up to 600 tok/s
- smallest smart MoE we have
- lots of world knowledge
- easy to run
—— 32-96gb ram/vram (2-10k usd)
- nex-n2-mini huggingface.co/nex-agi/Nex-N… builds on qwen3.6-35B and seems to do really well
- qwopus-27B huggingface.co/Jackrong/Qwop… this model topped a lot of our benchmarks at local.ai
—— 384gb vram (10-50K usd)
- huggingface.co/MiniMaxAI/Min… 23B means it’s close to qwen3.6-27B per token, while also have a lot of specialisation.
- fast inference
- top open weight model on AA
—— 768gb-1TB
- huggingface.co/moonshotai/Ki…
Kimi has always been a top player here and their last model cuts speed and cost down by 30%
- great vision support
- first coder model by moonshot
———
Top models:
1. Qwen3.6-35B
2. Qwen3.6-27B
3. Step-3.7-Flash
4. Minimax-M3
5. Deepseek-v4-flash
———
Budget sweet spots:
#1 - 1K usd
Single 3090 / Mac mini / Intel arc b70 / AMD
- Qwen / Gemma
#2 - 5k usd
DGX Spark / Mac m5 max / 4x 3090
- qwen / Gemma step and deepseek flash
#3 - 12k usd
RTX Pro 6000 / Mac Ultra / 2x Spark / 8x 3090
Ds4-flash / step-3.7-Flash and above
#4 - 24k usd
2x 6000 / 2x Mac Ultra / 4x Spark / Mix
Same as above
#5 - 50k usd
4x 6000 / 4x Max Ultra / 12x Spark / 2 H100
Minimax-m3 / nex-n2-pro / step-3.7-flash
#6 - 100k usd
GB300 station / 8x 6000 / 4x H200 / Mix
GLM-5.2 / Kimi-K2.7
———
Let’s keep the Internet free thanks for reading
63
97
872
45,235
Open source AI must win
Jun 13
As a result of a US government directive, we are suspending access to Claude Fable 5 for all users. You can continue to use all other Claude models.
Here’s what this means for you:
Across Claude products, new sessions will run on your selected default model or Opus 4.8, and existing Fable 5 sessions will end with an error.
On the Claude Platform, requests to Fable 5 will also return an error. Please update your integrations to other Claude models.
We know this is a disruption to your workflows; we appreciate your patience and support.
6
9
68
4,676
VRMアバターと Three.js で簡単に3Dゲームを作り始められるスターターテンプレート「vrm-game-starter」を公開しました！
✅ Three.js (WebGPU)
✅ 物理エンジン不要のキャラコン・Foot IK
✅ .vrmをドロップするだけでアバター差し替え
✅ ゲーム内レベルエディタ付き
#threejs #MadeWithVRM #vroid
3
44
156
24,607
jin retweeted
Jun 10
The scary part about Anthorpic's Fable nerf is not that it refuses to answer biology or cryptography. It's that it foreshadows what's coming. A world where a couple companies decide what you can and cannot do. They're building a new ruling class and you're not in it...
463
1,307
13,884
730,004
👀
🔥 AI just found 21 zero-days in FFmpeg.
That’s the video library bundled inside many apps, tools, containers, and devices. Some bugs sat untouched for 15–20 years.
Google Chrome also dropped PATCHES for a record 429 vulnerabilities this week.
Read: thehackernews.com/2026/06/ai…
1
10
998
Same cost of tokens as the zcash audit, 21 new vulns. FFmpeg quietly used in every major tech platform, app, and media device in the world, practically anything A/V related
Coincidentally AI recently helped discover a vuln in the Linux kernel that sat untouched for 23 years too.
1
5
338
Supply chain attacks are now deploying rootkits 😳 fyi it’s gg once a rootkit is found, best practices dictate a complete format and reinstall of the computer
In case of Ironworm it’s more tedious; you gotta *carefully* rotate credentials and audit your git repos / workflows too. Btw remember shai halud dead man switch that deletes your home directory if it detected its credentials got revoked? 💀
Say you’re careful, how can you be so sure about your agent? It can get prompt injected by a webpage or file (see AI Agent Traps paper by Google Deepmind).
Time to airgap our dev machines to like a cloud VM to limit some blast radius.
⚠️ New "IronWorm" supply-chain attack: 30 npm packages from @ asteroiddao shipped a malicious Rust binary firing on preinstall.
It sweeps 86 env vars 20 credential files (AWS, GCP, Vault, npm, plus AI keys like Anthropic & OpenAI), hits Exodus wallets, hides behind an eBPF rootkit, and beacons over Tor. Self-propagates via npm Trusted Publishing OIDC, with backdated commits faked as claude/dependabot/renovate.
1
10
1,543
Do any crypto teams have access to Mythos?
Does it even matter?
One researcher with Opus 4.8 discovered a PROTOCOL exploit for a major chain in a few days.
AI cybersecurity apocalypse is here. All software has bugs. Projects that can respond quickly more likely to survive.
3
2
13
1,175
> These laws would regulate every device capable of production by forcing it to be analyzed live by AI algorithms, report what you are doing to the government in real time, and shut down your machine if they do not like what you are making
Fuck this nonsense permission to build
New York’s recent “ghost gun bill” is not just an attack on the Second Amendment.
We continue to highlight and fight back against this new wave of spyware laws because they target firearms, but their scope goes far beyond guns.
They are an attack on American industry as a whole.
These laws would regulate every device capable of production by forcing it to be analyzed live by AI algorithms, report what you are doing to the government in real time, and shut down your machine if they do not like what you are making.
Louis Rossmann breaks this down excellently in his new video. Full video linked below.
1
7
990
reminder that the united states of america is trying to jail a guy that wrote non-custodial code to help people keep their onchain activity free from surveillance
he did not launder money, did not have central control, his code was immutable. he did not commit any known crime. in fact, he did everything that fincen says and hired lawyers ahead of time just to make sure. didn't matter
he wrote code that individuals can keep their privacy. and they are trying to put him behind bars for this
the dutch have already convicted his cofounder
you can rug an entire country by releasing countless shitcoins, bomb whoever you want, but god forbid you write immutable code that keeps your finances private
Jun 3
I hate asking for money. But that’s how the system works.
Legal defense isn’t free, and I need the community’s help to keep fighting.
Please donate, share, or retweet → freeromanstorm.com
Until this case is dismissed, building DeFi - protocols, UIs, tooling carries real legal risk for everyone in this space. The outcome here affects all of us.
37
56
351
25,520
All the advancements of ai coding open standards compounding makes me think a golden age for immersive web is near
Jun 3
🚀 Major upgrades just landed in SuperSplat, the free and open source platform for 3D Gaussian splatting.
Here is a 24-MILLION-Gaussian scan streaming live to a web browser. Near instant load time. Solid 60 fps.
How? A new compute-based WebGPU renderer automatic LOD streaming. 🧵
1
1
6
602
jin retweeted
May 23
Casually rendering ~a hundred million triangles, in the browser.
All thanks to @sea3dformat's cool Nanite-style renderer for @threejs.
I've added GLB-to-meshlet and PBR rendering to it, very exciting to see where this is going!
45
184
2,028
204,340
jin retweeted
Apr 17
What do 3D printing, mesh networks, solar, and crypto have in common?
They all make one type of person very nervous: the one who charges you monthly for something you could own.
8
17
87
2,456
👀 Think it’s time to export my data and run self hosted git infra for safe measure
We are investigating unauthorized access to GitHub’s internal repositories. While we currently have no evidence of impact to customer information stored outside of GitHub’s internal repositories (such as our customers’ enterprises, organizations, and repositories), we are closely monitoring our infrastructure for follow-on activity.
6
761