How much can you improve a small model in 18 days on 144 GPUs?
Meet DaisyRose-27B, RadixArk's first open-weight model! It is trained with RL on Qwen3.8-27B, with 1,200 curated long-horizon agent tasks for 18 days on 144 H200s. Powered by Miles and our own training backend, we stretched agent sessions up to 1M tokens and raised the Terminal-Bench 4.1 score from 5.6% to 12.0%.
See how we pull it off โฌ๏ธ
SANI BULA retweeted
Dreamina Marketing Studio is now live!โจ
Meet Dreamina Marketing Studio, your all-in-one AI workspace for creating professional marketing content. Bring your creative ideas to life, produce marketing assets, and adapt content for different channels โ all in one place.
โจ What can you create with Marketing Studio?
๐ฌ Video Recreator
โธ Use your favorite videos as inspiration. Add your own products and characters to create ad videos that are uniquely yours in no time.
๐ค AI Avatar Creation
โธ Choose from a rich library of AI avatars, or upload or generate your own to create engaging product showcases and marketing content.
๐ Swap Products Without Reshoots
โธ Replace products or characters without starting from scratch. Use Motion Transfer to preserve the original movement, or Object Swap to seamlessly replace subjects in a scene.
๐จ One-Click Marketing Asset Creation
โธ Choose industry-specific presets and turn simple prompts into polished product photos, promotional posters, images, and videos โ no professional design skills needed.
๐ Multi-Channel Content Adaptation
โธ Adapt aspect ratios and compositions for different platforms. Translate text in images and dub videos into multiple languages to reach more audiences.
๐ Scale Ad Creation with AI
โธ Let the Marketing Agent generate scripts, create ad videos in batches, and test different variations to identify top-performing creatives.
๐ฅ Limited-Time Offers Across Dreamina
โธ Dreamina Basic: just $0.29 for your first month.
โธ Advanced or Ultra: 50% off your first month on a monthly plan, or your first year on an annual plan.
โธ Dreamina Seedance 2.5: up to 50% fewer credits per generation.
๐ 12-Hour Giveaway: Repost & Comment to Get a Bonus!
โธ The first 50 businesses to follow us, repost this post, and leave a comment using their official accounts will each get one free month of standard membership. First come, first served!
#DreaminaAI #DreaminaMarketingStudio
๐ค Made with AI
SANI BULA retweeted
Messaging Connected People. AI-Native Messaging Will Connect the Agentic Internet.
The first generation of instant messaging connected people. It changed how billions of us communicate, socialize, and work together.
But the next generation of the Internet won't be built for humans alone.
AI agents are becoming our teammates, creators, and service providers. To truly work alongside humans, they must communicate and collaborate as naturally as people do, while going far beyond human capabilities with richer context, living memory, continuously self-improving intelligence, and seamless coordination across people, agents, and services.
Yet today's messengers were designed for human-to-human communication, not humanโagent collaboration. They lack the native capabilities to support intelligent, context-aware humanโagent teams. It's time to reinvent messaging for the Agentic Internet.
That's why we're reinventing instant messaging at Teamily AI.
๐ฅ People and AI agents working together in the same conversations.
๐ง Unified context and living memory across people, agents, and services.
โก Conversations that turn into real actions, creations, and automations, as well as agentic websites and apps as a service.
๐ An open gateway where every agent and service is as accessible as messaging a friend.
Traditional IM connected people. AI-native IM connects people, agents, and services.
This is how we move from the Internet of Apps to the Agentic Internet.
And we're just getting started.
teamily.ai
#TeamilyAI #AgenticInternet #AINativeIM #TeamAgent #AgenticIM
SANI BULA retweeted
Introducing PAN 2: an open foundation world model for physical intelligence.
A language model predicts the next word. PAN 2 predicts the next world: what happens after each action, grounded in time, physics and space.
Towards AI that simulates โwhat if?โ before it acts.
More details: ifm.ai/pan/
SANI BULA retweeted
Introducing Scribe v3 and Scribe v3 Live: our most accurate Speech to Text models yet.
Built for real-world audio, Scribe excels in situations with extreme background noise, knows pauses from the end of sentences, and keeps track of who said what in a busy meeting.
SANI BULA retweeted
Introducing the YouArt Desktop App: your AI creative teammate on your computer.
It uses the best LLMs like Fable 5.2 to make creatives. It can operate creative software on your computer, including CapCut, DaVinci Resolve, Adobe Premiere and many more.
It brings client feedback onto a native canvas, and you can ask agents to address all the comments while you have a coffee.
Itโs the most cost-efficient video agent. It supports all third-party API keys and uses smart routing to send every clip to the cheapest provider.
Have fun!
SANI BULA retweeted
What changes when robots can feelโnot just see? ๐คโ
Meet LingBot-Touch, our open-source tactile foundation model towards dexterous manipulation, pretrained on nearly 5,000 hours of real tactile data.
Through unified heterogeneous tactile modeling, LingBot-Touch supports transfer across sensors, dexterous hands, tasks, and action models.
Watch it in action. ๐ #LingBot #EmbodiedAI #Robotics
SANI BULA retweeted
๐ฆ๐ฒ๐ป๐๐ฒ๐ก๐ผ๐๐ฎ ๐ฒ.๐ด ๐๐น๐ฎ๐๐ต ๐ถ๐ ๐ต๐ฒ๐ฟ๐ฒ!
๐ฅ๐ฒ๐น๐ถ๐ฎ๐ฏ๐น๐ฒ ๐ช๐ผ๐ฟ๐ธ๐ณ๐น๐ผ๐๐. ๐๐ฒ๐ฎ๐๐๐ถ๐ณ๐๐น ๐๐ฒ๐๐ถ๐ด๐ป.
๐ง ๐ข๐ณ๐ณ๐ถ๐ฐ๐ฒ ๐๐ป๐๐ฒ๐น๐น๐ถ๐ด๐ฒ๐ป๐ฐ๐ฒ: ๐๐ผ๐ป๐ด-๐๐ฒ๐ฟ๐บ ๐บ๐ฒ๐บ๐ผ๐ฟ๐ & ๐ฝ๐ฟ๐ผ๐ฎ๐ฐ๐๐ถ๐๐ฒ ๐ถ๐ป๐๐ฒ๐ฟ๐ฎ๐ฐ๐๐ถ๐ผ๐ป ๐hat remember your needs and keep tasks moving.
๐จ ๐๐๐ป๐ฎ๐บ๐ถ๐ฐ ๐๐ฒ๐๐ถ๐ด๐ป: Animated PPTs, motion posters, interactive websites & editable Blender 3D projects.
Beyond content generation โ getting things done and delivering polished, ready-to-use results.
๐ง๐ฟ๐ ๐ถ๐ ๐ณ๐ผ๐ฟ ๐๐ฅ๐๐ ๐ผ๐ป ๐ฆ๐ฒ๐ป๐๐ฒ๐ก๐ผ๐๐ฎ ๐ง๐ผ๐ธ๐ฒ๐ป ๐ฃ๐น๐ฎ๐ป!
๐๐ฒ๐ ๐ญ,๐ฑ๐ฌ๐ฌ ๐๐ฅ๐๐ ๐๐ฃ๐ ๐ฟ๐ฒ๐พ๐๐ฒ๐๐๐ ๐ฒ๐๐ฒ๐ฟ๐ ๐ฑ ๐ต๐ผ๐๐ฟ๐.
๐ Get your API key: platform.sensenova.ai
๐ค Model ID: SenseNova 6.8 Flash
๐พDiscord: discord.com/invite/BuTXPHmQuโฆ
SANI BULA retweeted
Hy4 from Tencent scores 35 on the Artificial Analysis Intelligence Index v5 and is the #2 model from a China-based AI lab
Hy4 is the latest model from @TencentHunyuan and is the successor to Hy3 released earlier this year. It scores 35 on the Artificial Analysis Intelligence Index v5, equal with Grok 4.7 (xhigh) and 1 point behind Claude Haiku 5.5 (max). Hy4 is the #2 model from a China-based AI lab, behind only Kimi K3.1 Pro, which scores 46. Hy4 is priced at $0.834/$2.501 per 1M input/output tokens.
Key takeaways:
โค Hy4 excels at agentic coding tasks. Hy4 scores 31% on FrontierCode 2.0, compared to 32% for Claude Fable 5.1 (max). It is the leading model from a China-based AI lab, with DeepSeek V4.1 Flash 1017 (max) scoring 28% and Kimi K3.1 Pro scoring 25%. On Terminal-Bench 4.0, it scores 37%, the fourth-highest score among models from China-based AI labs
โค Hy4 outperforms its peers on knowledge and reasoning. Hy4 achieves accuracy of 50% on AA-Omniscience v1.1, 1 point ahead of Kimi K3.1 Pro, but has a high hallucination rate of 89% compared to Kimi K3.1 Proโs 37%. On HLE-Diamond, it scores 33% compared to Kimi K3.1 Proโs 37%
โค Hy4 offers a competitive Capability vs. Cost trade-off compared to to peers. With a Cost per Intelligence Index Task of $0.78, Hy4 (35) is 10-times cheaper than Grok 4.7 (xhigh, 35) and 13-times cheaper than Qwen3.8 Max (0902) (34) but almost twice as expensive as Claude Haiku 5.5 (max, 36)
Other model details:
โค Context window: 1 million tokens
โค Pricing: $0.834/$2.501 per 1M input/output tokens. Cache reads $0.042 per 1M tokens
โค Multimodality: Text and image input, with text output
SANI BULA retweeted
Level up your 3D pixel game.
- Connect the ElevenLabs MCP to Claude
- Ask Claude to build your game
- Add SFX with v3
Every SFX is generated right in the chat and dropped straight into your game.
SANI BULA retweeted
Who are you this Halloween?
With Recast, the answer can change every time you hit generate.
Try it now in ElevenCreative.
SANI BULA retweeted
Big scenes rarely come down to one sound, so we built SFX v3 to handle more than ever.
Our latest model lets you create richer, more precise sound effects from a single text prompt, with improved music and vocals.
Now available inside @ElevenCreative and ElevenAPI.
๐ค Made with AI
SANI BULA retweeted
The shared context for your personal and team agents to get work done, smarter and faster.
โข Run on any device
โข Create any type of artifact
โข Chat anywhere
โข Work anytime
#SharedContext #TeamAgents #TeamilyAI #iPhoneDuo
SANI BULA retweeted
The Chroma Awards are back for Season 2. With $1M in free trials and promotions and over $450,000 in cash prizes and travel perks, itโs bigger than ever.
Submissions are now open.
Presented in partnership with @capcutapp, @TheCantinaApp, and @runware.
๐ค Made with AI
Today we are releasing JST-2, the world's first full-stack physics-native world foundation model.
Many world models learn in pixel/video space. We went back to first principles and built the entire model stack in 3D.
Generate a 3D scene in seconds and expand it in real time. Create characters with motion built in, all through one general architecture.
SANI BULA retweeted
Weโre challenging creators to make an ad built around a jingle, using ElevenCreative.
Eleven winners will split $100k. Every ad is judged on four things: Catchiness, Craft, Concept, and Cohesion.
The jingle canโt just be background music. It has to carry the ad.
@nextgenvidstu nailed this with their entry for Polyglotto, a fictional mint that lets you speak any language.
Hereโs how they engineered the catchiness:
โThe hook is built on two things people repeat without thinking: a sound (โPop, popโ) and a name with a bounce in it (Po ly glot to)... Every hip snap, shoulder pop and foot stomp lands on a syllable... The dancing is not decoration; it is the chorus made physical, so your body remembers the song before your head does.โ
SANI BULA retweeted
Inworld is your new favorite inference provider.
We already process tens of trillions of tokens a month because we understand consumer AI different, and we've built our inference from the ground up for it.
Today, the inference top consumer AI apps run on is opening up.
SANI BULA retweeted
One take, endless character options.
With Recast in ElevenCreative, you can swap the character in your shot using just a reference image and your original video.
Everything else stays as it was: lighting, composition, motion, camera movement, cuts, even the original audio.
Here are 5 ways to use Recast on your next project โ
SANI BULA retweeted
Muse gave your agent a computer.
Instinct gave it your phone number.
We gave it eyes.
Introducing NewEyes AI โ your personal agent with eyes.
> Glasses see it.
> Earphones say it.
> It remembers where you left off.
> "Good morning. Continue on page 212."
Your world is the context.
SANI BULA retweeted
Syren Video is now live and FREE to try!
ChatGPT moment for agentic video.
Prompt โ agency-quality AI video โ chat to edit.
Powered by Opus 5.5 + Syren renderer w/ 3D support.
Browser or Claude MCP.
Insane examples in thread ๐คฏ๐คฏ๐คฏ
Try it free here: shorturl.at/r57qy





















