Hatice Ozen retweeted
🟢 The Ultimate Network, Coming Soon…
A powerful collaboration between YTU Startup House and @nvidia Together, we're opening the door to a new era in AI engineering.
Details are on the way, stay tuned.
#YTUStartupHouse #ComingSoon #AIEngineer #NVIDIA
i’ve largely disappeared since the transition to nvidia but have to log in just to retweet this⚡️
Artificial Analysis has measured 3,431 tokens/s on Gemma 4 31B via a private demonstration endpoint for NVIDIA Groq 3 LPX - the highest 100k-context output speed we have measured for Gemma 4 31B on either a private or public endpoint
@nvidia announced today that the Groq 3 LPX ‘interactive AI inference’ rack is in full-scale production and will enter operation later this year. NVIDIA granted us early access to a private deployment of the rack serving Gemma 4 31B for benchmarking purposes
We ran the standard 1k, 10k and 100k input sequence length prompts that we use for benchmarking serverless API providers. At both the 10k and 100k input sequence lengths, the endpoint delivered a median output speed of approx. 3,400 tokens/s, measured over 50 sequential (single-concurrency) requests. Output speed was maintained between the 10k and 100k input sequence lengths tested, indicating robust long-context inference performance
Congratulations to the @nvidia team on the launch!
Hatice Ozen retweeted
Join us for a session on the real-world side of AI development.
🎤 Hatice Özen (@ozenhati) (DevRel @NVIDIA) NeMo, generative AI workflows, open-source, and deploying AI in production.
🔗 Register: luma.com/fa9ai4oi
Only felt right that my first meal back in SF as an @nvidia employee was at Denny’s
some closing thoughts as i reflect on 2025:
in just 52 weeks, we've made breakthroughs in model intelligence and open-source went from catching up to leading -- so many labs, old and new, dropping releases every other week.
models uploaded to huggingface nearly doubled year over year thanks to open model progress that allowed for better bases for finetunes. 12,305 with "2024" in the name vs 23,794 with "2025." tried to get the real count via API and got rate limited, which goes to show all the HF activity in 2025 🤗 (@ClementDelangue would love a release date/year filter to easily see YOY open-source growth).
at groq, we shipped 40+ models this past year on our free and paid tiers to explore improved capabilities with you. between insane releases, feature shipping, keeping up with our developer community, and being chronically online, it's been a wild ride.
and instead of things winding down in Q4, we kept waking up to big moves. anthropic acquired bun. meta just acquired manus. among other things.
i think all of this is leading to 2026 being the year of the harness, increased compute, and robotics. models are good enough now (opus 4.5 continues to amaze me), so we'll see much more focus on good tooling -- CLIs, IDEs, agents. the infra layer on top of models as well as the infra underneath. LPUs, TPUs, GPUs. and, of course, this intelligence sewn into robotics -- already seeing all the reachy minis on my feed and can't wait to see what's built with them! need one for myself ASAP.
hard to believe everything we witnessed and worked on in 2025! excited for what's coming. happy new year! 🎉
meta’s 2026 revival has commenced
Manus is entering the next chapter: we’re joining forces with Meta to take general agents to the next level.
Full story on our blog: manus.im/blog/manus-joins-me…
had early access over the past few days and all i'll say for now is try m2.1 with @droid.
MiniMax M2.1 is officially live🚀
Built for real-world coding and AI-native organizations — from vibe builds to serious workflows.
A SOTA 10B-activated OSS coding & agent model, scoring 72.5% on SWE-multilingual and 88.6% on our newly open-sourced VIBE-bench, exceeding leading closed-source models like Gemini 3 Pro and Claude 4.5 Sonnet.
The most powerful OSS model for the agentic era is here.
i've personally been using @minimax_ai's minimax-m2 for coding for the past 2 weeks and it's a gem. keep your eyes open for open models. 🤗
Open models year in review
What a year! We're back with an updated open model builder tier list, our top models of the year, and our predictions for 2026.
First, the winning models:
1. DeepSeek R1 (@deepseek_ai): Transformed the AI world
2. Qwen 3 Family (@AlibabaGroup): The new default open models
3. Kimi K2 Family (@Kimi_Moonshot): Models that convinced the world that DeepSeek wasn't special and China would produce numerous leading models.
Runner up models: MiniMax M2 (@minimax_ai), GLM 4.5 (@Zai_org), GPT-OSS (@OpenAI), Gemma 3 (@GoogleAI), Olmo 3 (@allen_ai)
Honorable Mentions: Nvidia's (@nvidia) Parakeet speech-to-text model & Nemotron 2 LLM, Moondream 3 VLM (@moondreamai), Granite 4 LLMs (@IBMResearch), and HuggingFace's (@huggingface) SmolLM3.
Updated Tier list:
Frontier open labs: DeepSeek (@deepseek_ai), Qwen (@AlibabaGroup), and Kimi Moonshot (@Kimi_Moonshot)
Close behind: Z.ai (@Zai_org) & MiniMax AI (@minimax_ai) (notably none from the U.S. here and up)
Noteworthy (a mix of US & China): StepFun AI (@StepFun_ai), Ant Group's (@AntGroup/ @TheInclusionAI Inclusion AI, Meituan (@Meituan_LongCat), Tencent (@TencentHunyuan), IBM (@IBMResearch), Nvidia (@nvidia), Google (@GoogleAI), & Mistral (@MistralAI)
Then a bunch more below that, which we detail.
Predictions for 2026:
1. Scaling will continue with open models.
2. No substantive changes in the open model safety narrative.
3. Participation will continue to grow.
4. Ongoing general trends will continue w/ MoEs, hybrid attention, dense for fine-tuning.
5. The open and closed frontier gap will stay roughly the same on any public benchmarks.
6. No Llama-branded open model releases from Meta in 2026.
Read the full post on @interconnectsai -- link below.
