English
Konstantin Gladych
764 posts
Konstantin Gladych
@gladkos
Building @AtomicBot_ai Local Ai Agent 🦞 @atomic_chat_hq Local AI Models Engine previously @AtomicWallet 15M users
Katılım Haziran 2009
528 Takip Edilen7.2K Takipçiler
Konstantin Gladych retweetledi
We released iOS app for Hermes Agent 📱
Connect to your self-hosted agent over Tailscale, Cloudflare Tunnel or ngrok. Or deploy on a VPS. Run tasks and manage your agent from anywhere.
English
@ildunari @atomicbot_ai @testingcatalog @uzairansar Appreciate this! Any questions or feedback you are very welcome!
English
English
Atomic Bot released an iOS app for Hermes Agent, bringing mobile control to a self-hosted agent running 24/7 on your own VPS.
> Remote access via Tailscale, Cloudflare, or ngrok
> Open source and private by default
atomicbot.ai@atomicbot_ai
We released iOS app for Hermes Agent 📱 Connect to your self-hosted agent over Tailscale, Cloudflare Tunnel or ngrok. Or deploy on a VPS. Run tasks and manage your agent from anywhere.
English
@alexocheema Depends on the task. For coding would prefer dense qwen 27b. For simple tasks 35b moe performs better. DS4 outperforms all them, but very slow on consumers devices. Nemotron is a joke.
English
What model do you like best on 128GB local inference devices e.g. MacBook Pro / DGX Spark / Strix Halo?
Comment below if you're using a model not listed here.
English
While my Tesla is in a shop (teenage fender benders), I thought oh cool maybe I can get myself a nice 911 or a Ferrari rental. I opened Turo and they’re like $4-5k per week. For a car that doesn’t even drive itself.
English
@LyalinDotCom Great article, would like you to explore more models!
English
@bridgemindai 35b is moe model. Uses only 3b parameters at once, not the full 35. Dense 27b will give you about 30-40tps. And you don’t need 128gb. 48-64 is actually enough with background tasks for midsize models.
English
MacBook Pro M5 Max is fully set up and running local models.
I have never seen speeds this fast.
Qwen 3.6 35B and Gemma 4 31B are running blazing fast on 128GB of unified memory.
Faster than both of my stacked NVIDIA DGX Sparks sitting right next to it.
Initial impressions: Apple silicon in 2026 is no joke for local inference.
The M5 Max handles 30B+ parameter models like they're nothing.
Full review and comparison video coming soon.
English
English
English
@catalinmpit You can try Hermes VPS hosting on @atomicbot_ai super easy setup, one click connection to all these services and messengers.
English
I want to do the following with Hermes on Hetzner VPS:
1. Have a chat interface that I can access from everywhere (probably Telegram or Discord or even the bad WhatsApp).
2. Connect it to my tools and apps like:
- GitHub
- Google Calendar
- Email
- Notion
- Google Drive
3. Have it book calendar slots, categorize/reply/delete my emails, make code changes, organize my company's finances in Notion, create tasks, generate videos/images/voices/etc. and other stuff.
That's all I can think of by now.
English
English
Yesterday I went to @fdotinc Canopy Festival, thanks 🙏🏼
I met with @siliconmania in person 🔥
I also met tons of epic fresh companies doing incredible stuff.
This builder energy can almost only be found in San Francisco.
English
@shengkun_ye @fdotinc Was a pleasure to meet you guys! I’m those guy who’s building local ai agents 😉
English
six weeks ago, i quit my job.
then things started moving crazy fast.
we got into @fdotinc in our second week.
started launching every day.
got our first paying customer.
our paying customers keep topping up.
got our first check, then the second.
every day there’s so much going on. all your energy and emotions get compressed. before you can react, the next thing hits.
there are ups and downs. sometimes it’s exhausting. but i love it.
i love that life can hold so much more in the same amount of time. and i love being a founder.
English
Qwen 3.7-max beats Opus 4.7 and GPT-5.5
We tested three frontier models on a real agentic task: write a Tetris bot that plays the game and trains itself. Each model could read its own code, run benchmarks, and rewrite itself across 10 iterations. Then we compared the final bots head to head.
Qwen 3.7-Max: training cost $1.32, bot improvement +56%
Claude Opus 4.7: training cost $12.15, bot improvement +28%
GPT-5.5: training cost $2.85, bot improvement +7%
Qwen won on every dimension - biggest jump, 9× cheaper than Claude, 2× cheaper than GPT. Long agentic loops is where Qwen Max actually delivers.
English
Anthropic opening an office in Milan on 28th May was not on my bingo card 👀
English
@YangSu2000 Congrats! And we did a fun test x.com/atomic_chat_hq…
atomic.chat@atomic_chat_hq
Qwen 3.7-max beats Opus 4.7 and GPT-5.5 We tested three frontier models on a real agentic task: write a Tetris bot that plays the game and trains itself. Each model could read its own code, run benchmarks, and rewrite itself across 10 iterations. Then we compared the final bots head to head. Qwen 3.7-Max: training cost $1.32, bot improvement +56% Claude Opus 4.7: training cost $12.15, bot improvement +28% GPT-5.5: training cost $2.85, bot improvement +7% Qwen won on every dimension - biggest jump, 9× cheaper than Claude, 2× cheaper than GPT. Long agentic loops is where Qwen Max actually delivers.
English
We are excited to release Qwen3.7, featuring an extended version of our environment scaling approach. To maximize generalization across diverse downstream agentic tasks, we deliberately restricted ourselves from prior knowledge of many of the evaluations, which were not selected until pre-release, making them fully blind tests on out-of-domain environments.
The resulting scaling behavior is notably predictable: performance gains across any subset of benchmarks reliably correlate with gains on the rest, consistent with genuine capability generalization. Details in the upcoming technical report.
#qwen
Qwen@Alibaba_Qwen
Agent Scaling：Building on Qwen3.5's environment scaling approach, we've aggressively expanded the quality and diversity of agentic training environments in Qwen3.7 — agentic capabilities generalize from diverse environments, just as language models do from diverse text. The figure below shows a clear and consistent improvement trajectory, with Qwen3.7-Max achieving a top-3 average ranking that approaches Claude-4.6-Opus-Max.
English
@ChujieZheng Fun test: Qwen beats Opus and GPT in agentic tasks!
x.com/atomic_chat_hq…
atomic.chat@atomic_chat_hq
Qwen 3.7-max beats Opus 4.7 and GPT-5.5 We tested three frontier models on a real agentic task: write a Tetris bot that plays the game and trains itself. Each model could read its own code, run benchmarks, and rewrite itself across 10 iterations. Then we compared the final bots head to head. Qwen 3.7-Max: training cost $1.32, bot improvement +56% Claude Opus 4.7: training cost $12.15, bot improvement +28% GPT-5.5: training cost $2.85, bot improvement +7% Qwen won on every dimension - biggest jump, 9× cheaper than Claude, 2× cheaper than GPT. Long agentic loops is where Qwen Max actually delivers.
English
Qwen3.7-Max is our best flagship model built for real-world tasks and production environments. Try it and hope you enjoy!
Qwen@Alibaba_Qwen
📣Meet Qwen3.7-Max — our latest flagship, made for the Agent Era. A versatile foundation for agents that actually get things done: 🧑💻 Coding agent, end to end. Frontend prototypes, multi-file refactors, real debugging — nails it. 🗂️ A reliable office and productivity assistant. Get your work done through MCP integrations and multi-agent orchestration. ⏱️ Long-horizon autonomy. 35 hours straight on a kernel optimization task — 1,000+ tool calls, zero hand-holding. 🔌 Scaffold-agnostic. Claude Code, OpenClaw, Qwen Code, or your own stack. Consistent reliability everywhere. API's up on Alibaba Model Studio. You can also take it for a spin on Qwen Studio. Go build something wild!🏃🏃♂️ 📖 Blog: qwen.ai/blog?id=qwen3.7 ✅ Qwen Studio: chat.qwen.ai/?models=qwen3.… ⚡️ API：modelstudio.console.alibabacloud.com/ap-southeast-1…
English
@atomic_chat_hq Great research, Alibaba strikes! @ChujieZheng congrats!
English
Konstantin Gladych retweetledi
