San Francisco Bay Area
Joined May 2020
- Tweets1.7K
- Following696
- Followers254
- Likes286
Raphael Amorim retweeted
🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.
🔹 Introducing the smallest model in our new architecture family, with native visual understanding.
🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models.
1/6
Raphael Amorim retweeted
GLM 5.3 Flash is 0x Alpha for those that are curious
Tomorrow it's going to be a great day for Local AI with Qwen and GLM releasing new model weights.
Raphael Amorim retweeted
🚀 DeepSeek-V4-Flash Official API is now LIVE in public beta!
🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇
🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex!
Check out the configuration details in our official API docs: api-docs.deepseek.com/quick_…
Raphael Amorim retweeted
Inkling-small is out today! With SGLang, you can get 648 tok/s decode with DSpark (simulated acc len=4) and 288 tok/s w/o DSpark, under the same setup (8x @NVIDIAAI B200, TP 8, NVFP4, bs=1).
What makes this model different is the size. 276B total with 12B active is a sweet spot for RL, and both LoRA and full-parameter training become well within reach. Miles is ready and verified for multimodal RL on Inkling-small, so you can turn your multimodal data into real capability gains.
At ~1/4 the size, Inkling-small matches the bigger version in capability and even wins on some benchmarks.
Run Inkling-small with SGLang, and customize it with Miles.
Today, we are releasing Inkling-Small.
Inkling-Small achieves comparable performance to Inkling at a quarter of its size. It features 276B total parameters, 12B active. We are making the full weights available.
thinkingmachines.ai/news/ink…
Fine-tune it on Tinker today, or chat with it in text, image, and audio on Tinker Playground.
Raphael Amorim retweeted
Join us this Tuesday to learn more on Local AI, from software to hardware 🤗
We'll be joined by @TheAhmadOsman & @MikeBradleyAI covering hardware setups & local inference with live demo, and @alexocheema & @0xSero on picking model for your hardware, model compression and REAPs 🔥
Set your reminders to not miss out! 🔔
Raphael Amorim retweeted
Qwen3.8 is launching and going open-weight soon!🌐
With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5.
You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out.
Can't wait to hear what you build. Stay tuned! 🚀
Token Plan
international：qwencloud.com/pricing/token-…
China：platform.qianwenai.com/prici…
Raphael Amorim retweeted
Setting up with @NVIDIAAI
Local AI Summit, Room 2009 at @aiDotEngineer
Demos:
- Running GLM 5.2 on DGX Station
- Running Nemotron 3 Ultra on 4 x DGX Spark
Raphael Amorim retweeted
Check google/gemma-4-12B-it running on NVIDIA DGX Spark with vLLM! Thanks Rajendra Rawat for your contribution.
Run it with sparkrun.dev: sparkrun run @spark-arena/f1cbb973-cde2-4d0e-9ebf-538ece1993e9
View full benchmark at spark-arena.com/benchmark/f1…
Raphael Amorim retweeted
📣 @SKhynix and @NVIDIA announce a multiyear technology partnership to codevelop next-generation memory for the global AI factory buildout.
SK hynix will codevelop memory for NVIDIA's platforms — from NVIDIA Vera Rubin to Jetson Thor — while advancing fab digital twins using @NVIDIAOmniverse libraries and applying NVIDIA CUDA-X and PhysicsNeMo to accelerate semiconductor design and manufacturing.
Read the press release: nvda.ws/4e43e0p
Raphael Amorim retweeted
Our Open Source projects crossed 2000 stars on Github and 10.000 hours of benchmarks and over 100 recipes on Spark/GB10 devices. Thanks to our great community for supporting us with your contributions, insights, critique and benchmarks. You’re the reason we were able to reach those numbers. Join the discord channel to engage with the community and maintainers discord.gg/GH5kRgv6ZD
Raphael Amorim retweeted
New tech blog post on 3-Node Mesh Networking on DGX Spark
blog.spark-arena.com/posts/2…
With sparkrun.dev it's a 2 step process:
sparkrun cluster create mesh \
--hosts 10.0.0.11,10.0.0.12,10.0.0.13 \
--default
sparkrun setup wizard --cluster mesh
Raphael Amorim retweeted
we all live in a big ecosystem that all feeds into each other
- frontier labs push the bar on what's possible
- opensource labs build off of that and make their own discoveries
- inference providers make huge capex investments to serve these models
- app builders create real $$$ demand to justify continued investment
- individual hackers get these models running in all kinds of places
- lots of co-designing between these parties that improve things
- lots of conflict between these parties that create a good equilibrium
no one spot is morally superior to the others - very dumb to get caught up in that line of thinking
Raphael Amorim retweeted
🥈Qwen/Qwen3-Coder-Next-FP8 running on @sgl_project 2-node Spark cluster takes 2nd position on the leaderboard for decoding with 151.37 tokens/s on concurrency = 10
spark-arena.com/leaderboard