Konstantin Elstner @[email protected]

rotating frontend dev & nerd ... in love with all kind of devices, which has displays ^.- #⃣ but also a terminal junkie

[Ahmad @TheAhmadOsman]

Local AI hardware = capacity × bandwidth × software stack

- Capacity tells you what fits
- Bandwidth tells you how hard the box can breathe
- The software stack tells you how much of the spec sheet you can actually cash out.

Hardware by Memory Bandwidth
- Mac Studio M3 Ultra: up to 512GB @ 819 GB/s
- RTX PRO 6000 Blackwell: 96GB @ 1792 GB/s
- RTX 5090: 32GB @ 1792 GB/s
- RTX 4090: 24GB @ 1008 GB/s
- RX 7900 XTX: 24GB @ 960 GB/s
- Radeon PRO W7900: 48GB @ 864 GB/s
- AMD Radeon AI PRO R9700: 32GB @ 640 GB/s
- Intel Arc Pro B65: 32GB @ ~608 GB/s
- Tenstorrent Wormhole n300: 24GB @ 576 GB/s
- Tenstorrent Blackhole p150: 32GB @ 512 GB/s + 800G
- MacBook Pro M5 Max: 460-614 GB/s
- MacBook Pro M5 Pro: 307 GB/s
- DGX Spark: 128GB @ 273 GB/s (coherent + CUDA)
- Mac mini M4 Pro: 273 GB/s
- Ryzen AI Max / Strix Halo: ~256 GB/s (~96GB usable GPU)
- MacBook Air M5: 153 GB/s
- Snapdragon X2 Elite: 152-228 GB/s
- Intel Lunar Lake: 136 GB/s
- Snapdragon X Elite: 135 GB/s
- Mac mini M4: 120 GB/s
- Arc Pro B60: 24GB @ ~456 GB/s

Verdict

- GPUs are still the bandwidth kings

- Apple wins: stupid amounts of memory, don’t want to shard across GPUs
- Apple loses: when raw tokens/sec & concurrency matter more

- DGX Spark: coherent memory + NVIDIA stack

- Strix Halo / Ryzen AI Max: first real x86 unified-memory contender

- Tenstorrent: fully OSS stack, excited to see this mature

Fitting ≠ serving

Even if it fits, you still pay for
- bandwidth during decode
- KV cache growth
- dequantization
- batching + concurrency
- scheduler quality
- framework overhead

The only mental model that matters:

1. What must fit?
2. What bandwidth tier do I need?
3. What software stack can actually deliver it?

In short:
- NVIDIA → fastest raw speed
- Apple Studio M3 Ultra → biggest one-box memory
- Strix Halo → first real x86 unified
- DGX Spark → coherent NVIDIA dev appliance
- AMD / Intel Arc → rising alternatives
- Tenstorrent → fully opensource stack

Do ask: “which bottleneck am I buying?”

Not: “which hardware is best?”

[Nous Research @NousResearch]

The next evolution of Hermes Agent is here!

Introducing Hermes Desktop: everything you love about Hermes, now native on your machine.

First demoed in Jensen's GTC keynote, it's now in public preview.

[Nous Research @NousResearch]

Hermes Agent v0.12.0 - “The Curator Release”

[Nous Research @NousResearch]

Full Release Notes: https://t.co/nnO54pB3a0

Run 'hermes update' to enjoy

[Ahmad @TheAhmadOsman]

https://t.co/7nk1OjRA3u

[Qwen @Alibaba_Qwen]

⚡ Meet Qwen3.6-35B-A3B：Now Open-Source！🚀🚀

A sparse MoE model, 35B total params, 3B active. Apache 2.0 license.

🔥 Agentic coding on par with models 10x its active size
📷 Strong multimodal perception and reasoning ability
🧠 Multimodal thinking + non-thinking modes

Efficient. Powerful. Versatile. Try it now👇

Blog：https://t.co/EXx5y466su
Qwen Studio：https://t.co/bg4tAU1p74
HuggingFace：https://t.co/w4pDX14DZS
ModelScope：https://t.co/SuRyLzdQiO
API（‘Qwen3.6-Flash’ on Model Studio）：Coming soon～ Stay tuned

[Nous Research @NousResearch]

Hermes Agent v0.8.0 is here.

Full changelog below ↓

[Konstantin Elstner @[email protected] @dashart]

Hermes + Gemma 4 on DGX Spark = great performance, broken tool calling. Native tool support conflicts with Hermes logic → ~80% failure rate. Promising combo, not production-ready yet.

[Rohan Paul @rohanpaul_ai]

🇨🇳 China’s top models are climbing very quickly and the gap to the best US closed or top-tier models are shrinking fast.

And China’s best open-source models have already overtaken the US.

Open-source models spread through downloads, fine-tuning, and on-prem deployment, so leadership there can translate into faster global adoption even without controlling the top closed models.

Chart from FT

ft .com/content/d9af562c-1d37-41b7-9aa7-a838dce3f571

[Konstantin Elstner @[email protected] @dashart]

@vatro_vrbanic Tokens per sec? And Quality per Token?

[Konstantin Elstner @[email protected] @dashart]

Classic Playbook move of the Tech giants ...
https://t.co/Bbd5VKhA6k

[Konstantin Elstner @[email protected] @dashart]

#Jetbrains #networkgate issue reports
https://t.co/sRDOas1eiO

[Konstantin Elstner @[email protected] @dashart]

JetBrains has known for over two months about macOS 15 networking issues. They've posted a support document that's hard to find and are closing related bug reports as "not our problem."
If JetBrains can no longer support macOS, they should stop selling these products.
#jetbrains

[PK_SOE @pk_soe]

‼️⭕️🌉 Achtung Vollsperrung!
Ab sofort Sperrung, Brücke über die Elbe #B172 in Bad #Schandau bis auf Widerruf wegen Einsturzgefahr bzw. Korrosion der Brücke!

☝️ℹ️ Sofortige Sperrung der Elbbrücke in Bad Schandau – Notwendige Maßnahme aufgrund von Längsrissen im Unterspannband

[Konstantin Elstner @[email protected] @dashart]

Wow, entgegen den letzten Andeutungen ist der Avanced Voice Mode von #OpenAI jetzt auch in Deutschland verfügbar 😍

[Sam Altman @sama]

rollout complete; live to 100% of chatgpt plus/team users now

[Min Choi @minchoi]

The Evals on human exams and ML benchmarks are showing that it's significantly outperforming GPT-4o.

[Konstantin Elstner @[email protected] @dashart]

OMG, Mailgun was bought by Sinch! Now you have to re-auth your email to access, but their server is on a spam blacklist (https://t.co/ouu7gZaHRb), so nothing arrives. Support just asks for a service key you can't get without account access! #mailgun #servicehell

[Konstantin Elstner @[email protected] @dashart]

Cdu Sachsen macht sich schon mal ready für Koalition mit BSW oder AfD #facepalm https://t.co/SWYt3uPhsU

[AdonisJS @adonisframework]

It looks like the latest release of TypeScript breaks any AdonisJS application using our ORM (Lucid) - and maybe more.

We recommend that you stay on version 5.4 for the moment.