@Yuchenj_UW It would've been stranger if they didn't train from a strong base model. People acting like Cursor is the first company to built on top of open-source 😂
English
Darin
139 posts
Darin
@darin_ver
DevRel @FeatherlessAI, AI Engineer, eu/acc
Katılım Ocak 2025
188 Takip Edilen73 Takipçiler
People dunk on Cursor like: “it’s just Kimi K2.5,” “look inside, it’s a Chinese model.”
There’s no shame in building on top of strong base models and doing your own post-training or RL (as long as you respect the license).
In most cases you don’t need to pretrain from scratch. I think the whole industry will shift toward more post-training and RL on Chinese open-source models.
That’s also part of why we’re seeing the biggest GPU shortage and H100 price spike right now.
Yuchen Jin@Yuchenj_UW
Cursor’s Composer 2 is likely built on Kimi K2.5. The model URL + tokenizer are strong signals. I love this direction: companies mid-train and post-train on top of OSS LLMs. Prediction: open-source model labs will monetize by taking a cut when others build on top of their models and scale to millions of real users. They will enforce this via licensing. That’s the flywheel. That’s how open-source AI thrives.
English
Yep, Composer 2 started from an open-source base! We will do full pretraining in the future.
Only ~1/4 of the compute spent on the final model came from the base, the rest is from our training. This is why evals are very different.
And yes, we are following the license through our inference partner terms.
Fynn@fynnso
was messing with the OpenAI base URL in Cursor and caught this accounts/anysphere/models/kimi-k2p5-rl-0317-s515-fast so composer 2 is just Kimi K2.5 with RL at least rename the model ID
English
Darin retweetledi
Darin retweetledi
I didn't want to comment on this but
I saw Anthropic employees who mentioned distilling one model isn't innovation, just read papers from these labs they have good stuff there?
also your lab is distilling protected IP in masses technically?
Anthropic@AnthropicAI
We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax. These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.
English
Darin retweetledi
By popular demand, Nanbeige4.1-3B is now available on Featherless.
👇
featherless.ai/models/Nanbeig…
Nanbeige@nanbeige
🚀 Announcing Nanbeige4.1-3B – our latest open-source 3B model mastering reasoning, preference alignment, & agent capabilities! Try it now: huggingface.co/Nanbeige/Nanbe… reddit.com/r/LocalLLaMA/c…
English
calling it now: all these agent coding IDEs and GUIs are a phase and will be short lived.
most devs will be using only a terminal in a few months.
San Francisco, CA 🇺🇸 English
Darin retweetledi
Belgium has so many goated ai researchers, might be one of the highest per capita, topped only by Switzerland.
This might come as a surprise to you mostly because, except for myself, Belgians don't typically rub anything into your face. They just let things be.
yi@agihippo
Nationalistic ego is a strange thing. I see in some smaller countries where there are multiple famous researchers / startup founders, things can get complex and messy really quickly. Thankfully in Singapore there is one and only one legendary agi hippo 🦛
English
Some exciting stuff happening in the open-source TTS space 👀
Qwen@Alibaba_Qwen
Qwen3-TTS is officially live. We’ve open-sourced the full family—VoiceDesign, CustomVoice, and Base—bringing high quality to the open community. - 5 models (0.6B & 1.8B) - Free-form voice design & cloning - Support for 10 languages - SOTA 12Hz tokenizer for high compression - Full fine-tuning support - SOTA performance We believe this is arguably the most disruptive release in open-source TTS yet. Go ahead, break it and build something cool. 🚀 Everything is out now—weights, code, and paper. Enjoy. 🧵 Github: github.com/QwenLM/Qwen3-T… Hugging Face: huggingface.co/collections/Qw… ModelScope: modelscope.cn/collections/Qw… Blog: qwen.ai/blog?id=qwen3t… Paper: github.com/QwenLM/Qwen3-T… Hugging Face Demo: huggingface.co/spaces/Qwen/Qw… ModelScope Demo: modelscope.cn/studios/Qwen/Q… API: alibabacloud.com/help/en/model-…
English
Darin retweetledi
🍻 Singapore! We're hosting an Open-Source AI Social during Singapore AI Research Week
📅 Jan 23 | 6-8pm
📍 Singapore (location on registration)
Free drinks, food, and good conversation with AI researchers, builders & founders.
We just hit 19.7k models deployed, come help us celebrate!
RSVP 👇
English
Gemma been cooking lately! Can't wait for Gemma 4 to come out next
Google DeepMind@GoogleDeepMind
We’re releasing TranslateGemma, a new family of open translation models with support for 55 languages. 🌐 Available in 4B, 12B, and 27B parameter sizes – they’re designed for efficiency without sacrificing quality.
English
Darin retweetledi
Today, we release LFM2.5, our most capable family of tiny on-device foundation models.
It’s built to power reliable on-device agentic applications: higher quality, lower latency, and broader modality support in the ~1B parameter class.
> LFM2.5 builds on our LFM2 device-optimized hybrid architecture
> Pretraining scaled from 10T → 28T tokens
> Expanded reinforcement learning post-training
> Higher ceilings for instruction following
🧵
English
Ok, back to shipping. What do you want to see from GDM in the next couple of months?
(Gemini, Nano Banana, AI Studio, Gemma, Veo, and other AI dev products)
English
Saturday well spent!
Huge props to @pollenrobotics and @huggingface, the whole build experience was smooth start to finish!
Is the next step connecting it to @FeatherlessAI 🤔?
English
Singapore needs more love, who should we interview there? 😃
Michael Bronstein@mmbronstein
NeurIPS 2025 papers per 1 Million People 1. Singapore – 64.51 2. Switzerland – 22.13 3. Israel – 11.17 4. UAE – 9.47 5. UK – 7.50 6. US – 7.44 7. Denmark – 7.37 8. Australia – 7.31 9. Canada – 6.93 10. South Korea – 5.78
English
Another 2 new models from Mistral and their new CLI 👀
Mistral AI@MistralAI
Introducing the Devstral 2 coding model family. Two sizes, both open source. Also, meet Mistral Vibe, a native CLI, enabling end-to-end automation. 🧵
English
Darin retweetledi
Featherless is proud to sponsor inference at the Nigeria National AI Hackathon 2025 Finals hosted by @DeepFunding 🇳🇬
12 top teams from across Nigeria's 6 regions competing Dec 6-9. The winning team goes on to represent Nigeria at the African National AI Contest.
Good luck to all the finalists this weekend!
English