Xtracker

Neuralink has solved through-dura electrode implantation!

This is a very big deal, as it greatly improves the safety and ease of interfacing with the brain.

For Charlie

Every seat in this giant arena that isn’t roped off for security is packed to the ceiling. Honored to be here.

All for Charlie Kirk.

😢

I regret some of my posts about President @realDonaldTrump last week. They went too far.

This is not ok

Success is uncertain, but entertainment is guaranteed! ✨

I have a good feeling about 2025

Guess who?

Must be a coincidence 🙄

They say red light helps you sleep better

Huge thank you to everyone who supported this platform and our mutual quest to support freedom of speech!

It is morning in America again

The future is gonna be fantastic

The people of America gave @realDonaldTrump a crystal clear mandate for change tonight

You are the media now

Let that sink in

🇺🇸🇺🇸The future is gonna be so 🔥 🇺🇸🇺🇸

Game, set and match

The tower has caught the rocket!!

I am willing to serve

Haters will say this is AI 🕺🕺

😬

Good

Absolutely

This is real

El burro sabe mas que Maduro

The people of Venezuela have had enough of this clown 🤡

Adios Dictatora Maduro

Wow, Google has a search ban on President Donald Trump!

Election interference?

High time for an AI fashion show

Texas

I fully endorse President Trump and hope for his rapid recovery

To an exciting & inspiring future!

I am unreasonably excited about self-driving. It will be the first technology in many decades to visibly terraform outdoor physical spaces and way of life. Less parked cars. Less parking lots. Much greater safety for people in and out of cars. Less noise pollution. More space

I took delivery of a beautiful new shiny HW4 Tesla Model X today, so I immediately took it out for an FSD test drive, a bit like I used to do almost daily for 5 years. Basically... I'm amazed - it drives really, really well, smooth, confident, noticeably better than what I'm used

I quite like the new DeepSeek-OCR paper. It's a good OCR model (maybe a bit worse than dots), and yes data collection etc., but anyway it doesn't matter.

The more interesting part for me (esp as a computer vision at heart who is temporarily masquerading as a natural language

My pleasure to come on Dwarkesh last week, I thought the questions and conversation were really good.

I re-watched the pod just now too. First of all, yes I know, and I'm sorry that I speak so fast :). It's to my detriment because sometimes my speaking thread out-executes my

TV in the 90s: you turn it on, you watch.

TV 2025:
- turn on, wait for it to load
- popup: TV wants to update, 1.5GB. No.
- scroll sideways, find prime video app or etc
- popup: now app wants to update, 500MB. No!!
- App launching... App loading…
- select account screen
- 🫠

Excited to release new repo: nanochat!
(it's among the most unhinged I've written).

Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,

Finally had a chance to listen through this pod with Sutton, which was interesting and amusing.

As background, Sutton's "The Bitter Lesson" has become a bit of biblical text in frontier LLM circles. Researchers routinely talk about and ask whether this or that approach or idea

"AI isn't replacing radiologists" good article

Expectation: rapid progress in image recognition AI will delete radiology jobs (e.g. as famously predicted by Geoff Hinton now almost a decade ago). Reality: radiology is doing great and is growing.

There are a lot of imo naive

I think congrats again to OpenAI for cooking with GPT-5 Pro. This is the third time I've struggled on something complex/gnarly for an hour on and off with CC, then 5 Pro goes off for 10 minutes and comes back with code that works out of the box. I had CC read the 5 Pro version

I get ~10 spam calls per day (various automated voicemails, "loan pre-approval" etc) and ~5 spam messages per day (usually phishing).

- I have AT&T Active Armor, all of the above still slips through.
- All of the above is always from new, unique numbers so blocking doesn't work.

I am (slowly) re-reading the Tolkien legendarium (of which Lord of the Rings is a small part). The whole body of work is so incredible and there's nothing else like it... it dilutes other worlds of fiction. Wait - your story doesn't have a comprehensive history/mythology spanning

I'm noticing that due to (I think?) a lot of benchmarkmaxxing on long horizon tasks, LLMs are becoming a little too agentic by default, a little beyond my average use case.

For example in coding, the models now tend to reason for a fairly long time, they have an inclination to

Love this! Supercharger, diner, … but really a kind of exhibit for the future. Plotting a road trip SF -> LA to charge Shadowfax

Knowledge makes the world so much more beautiful.

How to build a thriving open source community by writing code like bacteria do 🦠. Bacterial code (genomes) are:

- small (each line of code costs energy)
- modular (organized into groups of swappable operons)
- self-contained (easily "copy paste-able" via horizontal gene

The race for LLM "cognitive core" - a few billion param model that maximally sacrifices encyclopedic knowledge for capability. It lives always-on and by default on every computer as the kernel of LLM personal computing.
Its features are slowly crystalizing:

- Natively multimodal

+1 for "context engineering" over "prompt engineering".

People associate prompts with short task descriptions you'd give an LLM in your day-to-day use. When in every industrial-strength LLM app, context engineering is the delicate art and science of filling the context window

Nice - my AI startup school talk is now up! Chapters:

0:00 Imo fair to say that software is changing quite fundamentally again. LLMs are a new kind of computer, and you program them *in English*. Hence I think they are well deserving of a major version upgrade in terms of

Part 2 of this mystery. Spotted on reddit.
In my test not 100% reproducible but still quite reproducible.
🤔

My sleep scores during recent travel were in the 90s. Now back in SF I am consistently back down to 70s, 80s.

I am increasingly convinced that this is due to traffic noise from a nearby road/intersection where I live - every ~10min, a car, truck, bus, or motorcycle with a very

Making slides manually feels especially painful now that you know Cursor for slides should exist but doesn’t.

An attempt to explain (current) ChatGPT versions.

I still run into many, many people who don't know that:
- o3 is the obvious best thing for important/hard things. It is a reasoning model that is much stronger than 4o and if you are using ChatGPT professionally and not using o3

We're missing (at least one) major paradigm for LLM learning. Not sure what to call it, possibly it has a name - system prompt learning?

Pretraining is for knowledge.
Finetuning (SL/RL) is for habitual behavior.

Both of these involve a change in parameters but a lot of human

A major mistake I made in my undergrad is that I focused way too much on mathematical lens of computing - computability, decidability, asymptotic complexity etc. And too little on physical lens - energy/heat of state change, data locality, parallelism, computer architecture. The

Noticing myself adopting a certain rhythm in AI-assisted coding (i.e. code I actually and professionally care about, contrast to vibe code).

1. Stuff everything relevant into context (this can take a while in big projects. If the project is small enough just stuff everything

The reality of building web apps in 2025 is that it's a bit like assembling IKEA furniture. There's no "full-stack" product with batteries included, you have to piece together and configure many individual services:

- frontend / backend (e.g. React, Next.js, APIs)
- hosting

I just vibe coded a whole iOS app in Swift (without having programmed in Swift before, though I learned some in the process) and now ~1 hour later it's actually running on my physical phone. It was so ez... I had my hand held through the entire process. Very cool.

I wrote a quick new post on "Digital Hygiene".

Basically there are some no-brainer decisions you can make in your life to dramatically improve the privacy and security of your computing and this post goes over some of them. Blog post link in the reply, but copy pasting below

It's 2025 and most content is still written for humans instead of LLMs. 99.9% of attention is about to be LLM attention, not human attention.

E.g. 99% of libraries still have docs that basically render to some pretty .html static pages assuming a human will click through them.

New 2h11m YouTube video: How I Use LLMs

This video continues my general audience series. The last one focused on how LLMs are trained, so I wanted to follow up with a more practical guide of the entire LLM ecosystem, including lots of examples of use in my own life.

Chapters

This is interesting as a first large diffusion-based LLM.

Most of the LLMs you've been seeing are ~clones as far as the core modeling approach goes. They're all trained "autoregressively", i.e. predicting tokens from left to right. Diffusion is different - it doesn't go left to

Agency > Intelligence

I had this intuitively wrong for decades, I think due to a pervasive cultural veneration of intelligence, various entertainment/media, obsession with IQ etc. Agency is significantly more powerful and significantly more scarce. Are you hiring for agency? Are

I was given early access to Grok 3 earlier today, making me I think one of the first few who could run a quick vibe check.

Thinking
✅ First, Grok 3 clearly has an around state of the art thinking model ("Think" button) and did great out of the box on my Settler's of Catan

Part of the reason for my 3hr general audience LLM intro video is I hope to inspire others to make equivalents in their own domains of expertise, as I’d love to watch them.

New 3h31m video on YouTube:
"Deep Dive into LLMs like ChatGPT"

This is a general audience deep dive into the Large Language Model (LLM) AI technology that powers ChatGPT and related products. It is covers the full training stack of how the models are developed, along with mental

There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper

We have to take the LLMs to school.

When you open any textbook, you'll see three major types of information:

1. Background information / exposition. The meat of the textbook that explains concepts. As you attend over it, your brain is training on that data. This is equivalent

For friends of open source: imo the highest leverage thing you can do is help construct a high diversity of RL environments that help elicit LLM cognitive strategies. To build a gym of sorts. This is a highly parallelizable task, which favors a large community of collaborators.

"Move 37" is the word-of-day - it's when an AI, trained via the trial-and-error process of reinforcement learning, discovers actions that are new, surprising, and secretly brilliant even to expert humans. It is a magical, just slightly unnerving, emergent phenomenon only

I don't have too too much to add on top of this earlier post on V3 and I think it applies to R1 too (which is the more recent, thinking equivalent).

I will say that Deep Learning has a legendary ravenous appetite for compute, like no other algorithm that has ever been developed

I still do this most days and I think it works great. My morning brain (right after 1hr exercise and 1 coffee) is quite eager to work and I go directly to the one top priority item. The energy decreases over time and with every distracting item loaded into the context window.

DeepSeek (Chinese AI co) making it look easy today with an open weights release of a frontier-grade LLM trained on a joke of a budget (2048 GPUs for 2 months, $6M).

For reference, this level of capability is supposed to require clusters of closer to 16K GPUs, the ones being

AI video generation today. When I was back in school, the story of the field of computer graphics (and physically based rendering etc.) was that we will carefully study and model all the object/scene geometry, physics, rendering etc., and after 1000 PhDs and 50 SIGGRAPHs get

The most bullish AI capability I'm looking for is not whether it's able to solve PhD grade problems. It's whether you'd hire it as a junior intern.

Not "solve this theorem" but "get your slack set up, read these onboarding docs, do this task and let's check in next week".

"I love traveling the world" 😂
(I think I reference this meme a lot so)

Of ~200 books I've read, the few that stayed with me over time and I find myself often thinking back to or referring to, in ~random order:

All short stories by Ted Chiang, especially Exhalation, Division By Zero, Understand, The Story of Your Life, Liking What You See, The

The reality of the Turing test

People have too inflated sense of what it means to "ask an AI" about something. The AI are language models trained basically by imitation on data from human labelers. Instead of the mysticism of "asking an AI", think of it more as "asking the average data labeler" on the

My Gladiator 2 review.

By chance I happened to watch this with the music of Interstellar playing in the background. Incredible. Huge 👏 to the team at SpaceX!!

🔥DeepSeek Input Cache Price Drop!

Effective immediately, the price for input cache hits across the ENTIRE DeepSeek API series is reduced to just 1/10th of the original price! Build more efficiently for less.

📌Reminder: The DeepSeek-V4-Pro 75% OFF promotion is still active

🛠 Open Source Release

🔗 Model: https://t.co/kORJG3nCWN
🔗 Tech report: https://t.co/X8Wcqbhg5a
🔗 Key GPU kernels in TileLang & CUDA (use TileLang for rapid research prototyping!)

4/n

💻 API Update

🎉 Lower costs, same access!
💰 DeepSeek API prices drop 50%+, effective immediately.

🔹 For comparison testing, V3.1-Terminus remains available via a temporary API until Oct 15th, 2025, 15:59 (UTC Time). Details: https://t.co/3RNKA89gHR
🔹 Feedback welcome:

⚡️ Efficiency Gains

🤖 DSA achieves fine-grained sparse attention with minimal impact on output quality — boosting long-context performance & reducing compute cost.
📊 Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.

2/n

🚀 Introducing DeepSeek-V3.2-Exp — our latest experimental model!

✨ Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention(DSA) for faster, more efficient training & inference on long context.
👉 Now live on App, Web, and API.
💰 API prices cut by 50%+!

1/n

📊 DeepSeek-V3.1-Terminus delivers more stable & reliable outputs across benchmarks compared to the previous version.

👉 Available now on: App / Web / API
🔗 Open-source weights here: https://t.co/Jh4RudofKm

Thanks to everyone for your feedback. It drives us to keep improving

🚀 DeepSeek-V3.1 → DeepSeek-V3.1-Terminus
The latest update builds on V3.1’s strengths while addressing key user feedback.

✨ What’s improved?
🌐 Language consistency: fewer CN/EN mix-ups & no more random chars.
🤖 Agent upgrades: stronger Code Agent & Search Agent performance.

Pricing Changes 💳

🔹 New pricing starts & off-peak discounts end at Sep 5th, 2025, 16:00 (UTC Time)
🔹 Until then, APIs follow current pricing
📝 Pricing page: https://t.co/IyYitNzedg

5/5

Model Update 🤖

🔹 V3.1 Base: 840B tokens continued pretraining for long context extension on top of V3
🔹 Tokenizer & chat template updated — new tokenizer config: https://t.co/r3y717EVFp
🔗 V3.1 Base Open-source weights: https://t.co/5wlDui34hH
🔗 V3.1 Open-source weights:

Tools & Agents Upgrades 🧰

📈 Better results on SWE / Terminal-Bench
🔍 Stronger multi-step reasoning for complex search tasks
⚡️ Big gains in thinking efficiency

3/5

API Update ⚙️

🔹 deepseek-chat → non-thinking mode
🔹 deepseek-reasoner → thinking mode
🧵 128K context for both
🔌 Anthropic API format supported: https://t.co/DcWmJMA1CP
✅ Strict Function Calling supported in Beta API: https://t.co/jFhJQ4wyN3
🚀 More API resources, smoother

Introducing DeepSeek-V3.1: our first step toward the agent era! 🚀

🧠 Hybrid inference: Think & Non-Think — one model, two modes
⚡️ Faster thinking: DeepSeek-V3.1-Think reaches answers in less time vs. DeepSeek-R1-0528
🛠️ Stronger agent skills: Post-training boosts tool use and

🚀 DeepSeek-R1-0528 is here!

🔹 Improved benchmark performance
🔹 Enhanced front-end capabilities
🔹 Reduced hallucinations
🔹 Supports JSON output & function calling

✅ Try it now: https://t.co/IMbTch8Pii
🔌 No change to API usage — docs here: https://t.co/Qf97ASptDD
🔗

🚀 DeepSeek-V3-0324 is out now!

🔹 Major boost in reasoning performance
🔹 Stronger front-end development skills
🔹 Smarter tool-use capabilities

✅ For non-complex reasoning tasks, we recommend using V3 — just turn off “DeepThink”
🔌 API usage remains unchanged
📜 Models are

🚀 Day 6 of #OpenSourceWeek: One More Thing – DeepSeek-V3/R1 Inference System Overview

Optimized throughput and latency via:
🔧 Cross-node EP-powered batch scaling
🔄 Computation-communication overlap
⚖️ Load balancing

Statistics of DeepSeek's Online Service:
⚡ 73.7k/14.8k

🚀 Day 5 of #OpenSourceWeek: 3FS, Thruster for All DeepSeek Data Access

Fire-Flyer File System (3FS) - a parallel file system that utilizes the full bandwidth of modern SSDs and RDMA networks.

⚡ 6.6 TiB/s aggregate read throughput in a 180-node cluster
⚡ 3.66 TiB/min

🚀 Day 4 of #OpenSourceWeek: Optimized Parallelism Strategies

✅ DualPipe - a bidirectional pipeline parallelism algorithm for computation-communication overlap in V3/R1 training.
🔗 https://t.co/GBtxSvWLT4

✅ EPLB - an expert-parallel load balancer for V3/R1.
🔗

🚨 Off-Peak Discounts Alert!

Starting today, enjoy off-peak discounts on the DeepSeek API Platform from 16:30–00:30 UTC daily:

🔹 DeepSeek-V3 at 50% off
🔹 DeepSeek-R1 at a massive 75% off

Maximize your resources smarter — save more during these high-value hours!

🚀 Day 3 of #OpenSourceWeek: DeepGEMM

Introducing DeepGEMM - an FP8 GEMM library that supports both dense and MoE GEMMs, powering V3/R1 training and inference.

⚡ Up to 1350+ FP8 TFLOPS on Hopper GPUs
✅ No heavy dependency, as clean as a tutorial
✅ Fully Just-In-Time compiled

🚀 Day 2 of #OpenSourceWeek: DeepEP

Excited to introduce DeepEP - the first open-source EP communication library for MoE model training and inference.

✅ Efficient and optimized all-to-all communication
✅ Both intranode and internode support with NVLink and RDMA
✅

🚀 Day 1 of #OpenSourceWeek: FlashMLA

Honored to share FlashMLA - our efficient MLA decoding kernel for Hopper GPUs, optimized for variable-length sequences and now in production.

✅ BF16 support
✅ Paged KV cache (block size 64)
⚡ 3000 GB/s memory-bound & 580 TFLOPS

🚀 Day 0: Warming up for #OpenSourceWeek!

We're a tiny team @deepseek_ai exploring AGI. Starting next week, we'll be open-sourcing 5 repos, sharing our small but sincere progress with full transparency.

These humble building blocks in our online service have been documented,

🚀 Introducing NSA: A Hardware-Aligned and Natively Trainable Sparse Attention mechanism for ultra-fast long-context training & inference!

Core components of NSA:
• Dynamic hierarchical sparse strategy
• Coarse-grained token compression
• Fine-grained token selection

💡 With

🎉 Excited to see everyone’s enthusiasm for deploying DeepSeek-R1! Here are our recommended settings for the best experience:

• No system prompt
• Temperature: 0.6
• Official prompts for search & file upload: https://t.co/TtjEvldTz5
• Guidelines to mitigate model bypass

📢 Terminology Correction: DeepSeek-R1’s code and models are released under the MIT License.

@dapangdun This is a typical impersonation account. Please do not trust any information from this account.

To prevent any potential harm, we reiterate that @deepseek_ai is our sole official account on Twitter/X.

Any accounts:
- representing us
- using identical avatars
- using similar names
are impersonations.

Please stay vigilant to avoid being misled!

🌐 API Access & Pricing

⚙️ Use DeepSeek-R1 by setting model=deepseek-reasoner
💰 $0.14 / million input tokens (cache hit)
💰 $0.55 / million input tokens (cache miss)
💰 $2.19 / million output tokens

📖 API guide: https://t.co/Qf97ASptDD

🐋 5/n

🛠️ DeepSeek-R1: Technical Highlights

📈 Large-scale RL in post-training
🏆 Significant performance boost with minimal labeled data
🔢 Math, code, and reasoning tasks on par with OpenAI-o1
📄 More details: https://t.co/jWMxMVhGAQ

🐋 4/n

📜 License Update!

🔄 DeepSeek-R1 is now MIT licensed for clear open access
🔓 Open for the community to leverage model weights & outputs
🛠️ API outputs can now be used for fine-tuning & distillation

🐋 3/n

🔥 Bonus: Open-Source Distilled Models!

🔬 Distilled from DeepSeek-R1, 6 small models fully open-sourced
📏 32B & 70B models on par with OpenAI-o1-mini
🤝 Empowering the open-source community

🌍 Pushing the boundaries of **open AI**!

🐋 2/n

🚀 DeepSeek-R1 is here!

⚡ Performance on par with OpenAI-o1
📖 Fully open-source model & technical report
🏆 MIT licensed: Distill & commercialize freely!

🌐 Website & API are live now! Try DeepThink at https://t.co/v1TFy7LHNy today!

🐋 1/n

⚠️ Important Notice:

✅ 100% FREE - No ads, no in-app purchases
🛡️ Download only from official channels to avoid being misled
📲 Search "DeepSeek" in your app store or visit our website for direct links

🌟 3/3

✨ Key Features of DeepSeek App:

🔐 Easy login: E-mail/Google Account/Apple ID
☁️ Cross-platform chat history sync
🔍 Web search & Deep-Think mode
📄 File upload & text extraction

🌟 2/3

🎉 Introducing DeepSeek App!

💡 Powered by world-class DeepSeek-V3
🆓 FREE to use with seamless interaction
📱 Now officially available on App Store & Google Play & Major Android markets
🔗Download now: https://t.co/DIwqqkbK93

🌟 1/3

DeepSeek has not issued any cryptocurrency. Currently, there is only one official account on the Twitter platform. We will not contact anyone through other accounts.Please stay vigilant and guard against potential scams.

💰 API Pricing Update

🎉 Until Feb 8: same as V2!
🤯 From Feb 8 onwards:
Input: $0.27/million tokens ($0.07/million tokens with cache hits)
Output: $1.10/million tokens

🔥 Still the best value in the market!

🐋 3/n

🌌 Open-source spirit + Longtermism to inclusive AGI

🌟 DeepSeek’s mission is unwavering. We’re thrilled to share our progress with the community and see the gap between open and closed models narrowing.

🚀 This is just the beginning! Look forward to multimodal support and

🎉 What’s new in V3?

🧠 671B MoE parameters
🚀 37B activated parameters
📚 Trained on 14.8T high-quality tokens

🔗 Dive deeper here:
Model 👉 https://t.co/9iwEF6aLuk
Paper 👉 https://t.co/ruzwMFYAAH

🐋 2/n

🚀 Introducing DeepSeek-V3!

Biggest leap forward yet:
⚡ 60 tokens/second (3x faster than V2!)
💪 Enhanced capabilities
🛠 API compatibility intact
🌍 Fully open-source models & papers

🐋 1/n

🎉 What Else Can DeepSeek-VL2 Do?

😂 Understand memes like a pro!
🔍 Identify objects from image context!
📖 Craft stories from images—let it be your storyteller!

🧵 5/n

💼 Let DeepSeek-VL2 Boost Your Productivity!

🧠 Easily understand complex charts and diagrams
💻 Instantly generate plot code with precision

🧵 4/n

🔧 Technical Improvements

📊 Data: 2x high-quality training data vs. DeepSeek-VL1, unlocking new capabilities
🏗️ Architecture: Dynamic image tiling for flexible resolutions + efficient DeepSeek-MoE for language
⚙️ Training: Retains 3-stage training + new multi-modal parallel

💡 Still Fully Open-Source! Come explore our model and check out the code—technical report coming soon!

💾 Hugging Face: https://t.co/5CnlWKCq5t
💻 Github Page: https://t.co/2skukDeIjW

🧵 2/n

🎉 DeepSeek-VL2 is here! Our next-gen vision-language model enters the MoE era.

🤖 DeepSeek-MoE arch + dynamic image tilling
⚡ 3B/16B/27B sizes for flexible use
🏆 Outstanding performance across all benchmarks

🧵 1/n

🙌 With the release of DeepSeek-V2.5-1210, the V2 series comes to an end.
💪 Since May, the DeepSeek V2 series has brought 5 impactful updates, earning your trust and support along the way.
✨ As V2 closes, it’s not the end—it’s the beginning of something greater. DeepSeek is

📊 DeepSeek-V2.5-1210 raises the bar across benchmarks like math, coding, writing, and roleplay—built to serve all your work and life needs.

🔧 Explore the open-source model on Hugging Face: https://t.co/PaHvAE92ly

🧵(2/3)

🚀 DeepSeek-V2.5-1210: The Grand Finale 🎉

🌐 Internet Search is now live on the web! Visit https://t.co/IMbTch8Pii and toggle “Internet Search” for real-time answers. 🕒

🧵(1/3)

🌟 Inference Scaling Laws of DeepSeek-R1-Lite-Preview
Longer Reasoning, Better Performance. DeepSeek-R1-Lite-Preview shows steady score improvements on AIME as thought length increases.

🌟 Impressive Results of DeepSeek-R1-Lite-Preview Across Benchmarks!

The Kimi API is now live on AWS Marketplace. 🚀

If your team is already running on AWS, you can now access Kimi with consolidated billing. Plus, eligible customers can apply Kimi API usage directly toward their AWS EDP commitments.

Build and scale with Kimi today:

Introducing Goal Mode in Kimi Work

Goal lets your desktop agent run 24/7 until the task is done, built for long-horizon tasks and complex multi-step workflows.

RT @prz_chojecki: Kimi 2.7 ranked 2nd after Fable 5 and before GPT-5 xhigh

We have re-run our ErdosBench smoke test on 14 problems with Ki…

🌘 Meet Kimi K2.7 Code HighSpeed!
A high-speed mode of our latest open-source multimodal coding model, Kimi K2.7 Code.

⚡️ Up to 6× faster: Around 180 tok/s on coding tasks with median-length inputs, and up to 260 tok/s on shorter-context tasks.

🔷 Rolling out to Kimi Code Beta

Extra API quota for Kimi K2.7 Code builders 🎉

If you're building with Kimi API, get 20%–30% extra quota when you top up $100+ by July 2!

🔷 $100–$299 → +20% quota
🔷 $300–$999 → +25% quota
🔷 $1,000+ → +30% quota

(One bonus per account.)

- Details: https://t.co/HLRaFecpoN

RT @vllm_project: 🎉 Congrats to @Kimi_Moonshot on Kimi K2.7-Code, a coding-focused agentic model built on K2.6.

✨ 1T-parameter Mixture-of-…

🔗 Weights & code: https://t.co/DvAKMMcTjz

🎸 We're also launching the Kimi Code Beta Program today. Apply now if you'd like to try upcoming models and features before public release 👉 https://t.co/ArGpJcaYma

🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!

🔷 Improved coding & agent performance over K2.6: +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, and +31.5% on MLS Bench Lite.
🔷 Reasoning efficiency: Less overthinking, with 30% lower

🔗Try it now: https://t.co/MmfIaY72hP
We're just getting started. More data sources, more tools, more agent capabilities are coming soon!

Agent Swarm & Instant Document Delivery

Kimi will automatically coordinate 300 sub-agents to break down and execute your tasks. Delivers production-ready output in PPTX, Word, PDF, and Excel, straight to your desktop.

Built for Finance, with Data Source Tool Call

Kimi Desktop paired with agent swarm and data tool call, 300 specialized agents in parallel, each with access to tools:
• Yahoo Finance for market data
• World Bank for economic research
• Binance for crypto analysis
And more

Pair it with WebBridge and your agent will navigate websites in your browser: search, scroll, click, type and complete tasks.

Try it 👉 https://t.co/dFMx5atf2e

Meet Kimi Work - a local AI agent on your desktop that does the work for you.

🔹Native agent swarm: Up to 300 AI agents running in parallel on your local machine.
🔹Browser use: Paired with WebBridge extension, your agent will navigate websites in your browser: search, scroll,

RT @KimiDevs: Kimi Code, our open-source coding agent, just got a major upgrade!

🔹One-line CLI install, zero setup, fast startup​
🔹Drag in…

RT @Designarena: An open source model has returned to #1 on the 3D Design leaderboard by Design Arena.

Kimi K2.6 has reached the top of th…

RT @cerebras: Cerebras is now running Kimi K2.6 – a trillion parameter model – in enterprise trials.

At ~1,000 tokens/s, this is the fast…

RT @cursor_ai: Composer 2.5 is built on the same open-source base as Composer 2, Moonshot’s Kimi K2.5.

Supports Kimi Code CLI, Claude Code, Cursor, Codex, Hermes, and more.

Try it at: https://t.co/sUqDpi0HQr and the Chrome Web Store.

RT @opencode: steady growth for minimax m3

Day two of @aiDotEngineer started with a conversation anyone serious about open weights should be paying attention to.

@olive_jy_song, research lead RL, joined @Thom_Wolf to dig into sparse attention, native multimodal training from day zero, and why open-weights matter for

RT @pamelafox: Highlights from the @aiDotEngineer keynote chat with @olive_jy_song from @MiniMax_AI, creators of the open-weight M3 model:…

Our favorite vibe coder🧋

RT @VictorSuOrtiz: Missed our talk with @togethercompute on @MiniMax_AI sparse attention and kernel optimizations?

Catch us tomorrow at 10…

RT @brianchew: @olive_jy_song and @Thom_Wolf!!

big on @huggingface and @MiniMax_AI 🫶🏻🫶🏻🫶🏻

RT @ASI_LatAm: @SingularityNET @ASI_Alliance + @MiniMax_AI
Creando juntos con OmegaClaw y ASI:Create
https://t.co/0wC1g5w65Q

Finallyyy with @LambdaAPI

RT @VictorSuOrtiz: Here today lab #1 @aiDotEngineer for @MiniMax_AI

Come find us.

Catch us tomorrow at 1 PM UTC as we recap the inaugural BGI Sprint.

This is a glimpse of where local AI is heading and we are glad to be part of it.

Really impressive work by all the teams involved @Gradient_HQ, @tryParallax, and @GA_agent_ai

we’ll be at Lab #1 during @aiDotEngineer World’s Fair.

hope to see you there!

Congrats to all the winners of our cohosted hackathon with @cysic_xyz

Check out the stellar projects built with M3 👇

@aiDotEngineer @togethercompute @olive_jy_song @realDanFu Check it out here: https://t.co/Y33GG4C09d

next week at @aiDotEngineer, we are joining @togethercompute for a conversation on what goes into running agents at scale.

@olive_jy_song, Research Lead, RL at MiniMax, and
@realDanFu, VP of Kernels at Together AI, will walk through both sides of M3: the training decisions

RT @VictorSuOrtiz: Kicking off @aiDotEngineer week with @MiniMax_AI @Modular @digitalocean @GoogleDeepMind @livekit

RT @questflow: New free model on Questflow 🎁

MiniMax M3 is now live on the Questflow and in our AI Trader Arena.

Access is 100% free, ful…

👀 Looking forward to seeing builders give it a try tomorrow.

Curious what model is powering it, @browser_use

More great options for the open-weight ecosystem. Thanks @NVIDIAAI for making MiniMax M3 available in NVFP4.

@vercel @merge_api @FactoryAI @olive_jy_song Register here -> https://t.co/bsuGI24nYG

we’ll be at AI Engineer After Dark on July 1st with @vercel , @merge_api, @FactoryAI , and a room full of people building the AI engineering stack.

our Research Lead, RL Training @olive_jy_song will be giving a lightning talk on post-training MiniMax M3 as part of a lineup

RT @CreaoAI: Last night in SF, we gave a room full of creators one challenge: come with an idea, leave with a working demo.

Our CTO @intui…

Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities

- Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas
- MiniMax Sparse Attention scales context to 1M
-

for GLM-4.6, which features do you want most? Speed up to 100t/s? Stability? Lower price?

We need to try our best to close the gap

cooool

add oil

try 4.6 very soon

Great job

今天是1024程序员节，在此智谱为大家带来一个「为期 8 天」的小惊喜。

GLM Coding Plan 限时特惠，将SOTA的 Agentic Coding能力，带给更多个人/企业开发者：
- 新人首单 5 折；
- 成功邀请新人下单即可返 40% 平台赠金；
- 被邀好友通过链接下单还可获 10% 优惠。

how about https://t.co/VoJ7YSlnRR?

great

Glyph, a concurrent and comparable model to DeepSeek-OCR https://t.co/kZwc45Qtlz

Glyph: Scaling Context Windows via Visual-Text Compression

run in home

The best open model

hello world

GLM 4.6 #1 open model

cool

https://t.co/YzhgyhnjyX global

On webdev arena, GLM is ranked #4 and #1 in all open models

a picture from LMArena

GLM

Really?

cool

GLM写生产级代码

great

Who invented residual neural networks?

RL is probably one of the only things that can be easily published ...

glm+opencode

coool

less is more

nice

glm #1

Will do. Thanks a lot.

@aspenwokeley Thanks. AGI is all we need.

Cool

We are developing the next version of GLM. What features do you want the most?

great

coool

increadable

cool

self-evolution is the next step

Thanks for the support. Shall we have the model on Cerebras soon?

GLM-4.6 the highest score

@love_deepseek end of this year

We are #1 on the trending of OpenRouter today. Thanks for the support.

"GLM-4.6 - top of the charts, open-source"

GLM-4.6 is #1 in all open models!

Finally, our open-source GLM-4.6 are trending no. 1 on HF. Thanks to all for the support. We are working on the next version and stay tuned!

GLM

GLM

GLM

This might be the biggest AI hackathon ever:

* >6,300 registrants
* Runs for 2 weeks (Nov. 14-30)
* Open to anyone, anywhere virtually
* $20,000 in cash prizes + $3.5M+ in sponsor credits

Hosted by @Anthropic and @Gradio, along with 10 sponsors, join kickoff in 30 minutes 👇

Sun's out, models out. 😎
@IBM & @NASA dropped Surya, an open-source heliophysics model trained on 14 years of observations from NASA’s Solar Dynamics Observatory, and it's 🔥🔥🔥.

Welcome to @OpenAI on @huggingface!

Kimi K2 is number one trending on HF, congrats!

Super happy to announce that we are acquiring @pollenrobotics to bring open-source robots to the world! 🤖

Since @RemiCadene joined us from Tesla, we’ve become the most widely used software platform for open robotics thanks to @LeRobotHF and the Hugging Face Hub. Now, we’re

We are excited to partner with @AIatMeta to welcome Llama 4 Maverick (402B) & Scout (109B) natively multimodal Language Models on the Hugging Face Hub with Xet 🤗

Both MoE models trained on up-to 40 Trillion tokens, pre-trained on 200 languages and significantly outperforms its

@deepseek_ai Congratulations on the stellar release! 🤩

The model checkpoints and a detailed report - truly Christmas is here!

https://t.co/rzP6W128Xj

QwQ is #1 trending!

GoogleからGemma-2-JPNがリリースされました！このモデルはGemma 2 2Bを日本語でfine-tuneしたものです。Gemma 2の英語での性能と同レベルの性能で日本語をサポートします。
モデル一覧:

https://t.co/LgyHREpCOf

We passed 5 million users.

🥳That's 5 million of you who have signed up on the Hub 🚀 thank you for contributing to the ecosystem and making open Machine Learning happen!

We're just getting started 🤗

Which is your favorite open LLM? Why? 🤗

The Open Source community is amazing 🤗

We're having some infra issues; we're working on it. Please send hugs! 🤗

In the meantime,

import os
os.environ['HF_HUB_OFFLINE']=1

Hugging Face 🫶 @GoogleColab

With the latest release of huggingface_hub, you don't need to manually log in anymore. Create a secret once and share it with every notebook you run. 🤗

pip install --upgrade huggingface_hub

Check it out!👇

Zephyr 7b beats Llama 70b on MT Bench 🤯🤯🤯 https://t.co/MblJoRbKsR

How to train a Llama 2 chatbot (a step-by-step guide, designed for non-coders)

https://t.co/hBy35eihxE

Code Llama: Now on Hugging Chat 💻🦙

Try out the 34B Instruct model for free with super fast inference!

👉 https://t.co/TQ0ZaVZcdi

TRL 🤗 Hugging Face

Excited to announce that we're doubling down on our efforts to democratize RLHF and reinforcement learning with TRL, new addition to the @huggingface family, developed and led by team member @lvwerra 🎉🎉

Train your first RLHF model 👉https://t.co/sQRUPllJWT

Hugging Face is now part of the PyTorch Foundation as a premier member 🤝

We have been collaborating with the PyTorch team for the past four years and are committed to supporting the project.

We share an objective: to lower the barrier of entry to ML.

https://t.co/Sq3eL9xhVl

Llama 2: Now on Hugging Chat 🤗🦙

Try out the 70B Chat model for free with super fast inference, web search, and powered by open-source tools!

👉 https://t.co/TQ0ZaVZcdi

At Hugging Face, we are working to enable you to easily build and serve your own LLMs 🧑‍💻👨‍💻👩‍💻
In this blog, we talk about the amazing world of open-source LLMs, the challenges, and how the Hugging Face ecosystem can help you 🪐
Read about them here 👉https://t.co/cOgLmtBQS7

We are looking into an incident where a malicious user took control over the Hub organizations of Meta/Facebook & Intel via reused employee passwords that were compromised in a data breach on another site. We will keep you updated 🤗

📣 Calling all game dev and AI enthusiasts!🎮

Already 400 people signed up for the first Open Source AI Game Jam, where you'll use AI tools to make a game in a weekend🔥

Sign up here 👉 https://t.co/iDyTq9aXmA

What AI tools? Let's focus today on Audio tools 🔊
⬇️

🚨Exciting news! Next week, we’ll be launching a brand-new Audio Course! 🤗

Sign up today (https://t.co/2cLmirTddm) and join us for a LIVE course launch event featuring amazing guests like @DynamicWebPaige, Seokhwan Kim, and @functiontelechy! ⚡️

https://t.co/jy7Y2FIQgf

🤗 Transformers has been built by, with, and for the community.

Reaching 100k ⭐ on GitHub is a testament to ML's reach and the community's will to innovate and contribute.

To celebrate, we highlight 100 incredible projects in transformers' vicinity.

https://t.co/gMI1WGRhTm

The first RNN in transformers! 🤯
Announcing the integration of RWKV models in transformers with @BlinkDL_AI and RWKV community!
RWKV is an attention free model that combines the best from RNNs and transformers.
Learn more about the model in this blogpost: https://t.co/0FQmsaRVZw

We just released Transformers' boldest feature: Transformers Agents.

This removes the barrier of entry to machine learning

Control 100,000+ HF models by talking to Transformers and Diffusers

Fully multimodal agent: text, images, video, audio, docs...🌎

https://t.co/OILVxIX44I

SAM, the groundbreaking segmentation model from @Meta is now in available in 🤗 Transformers!
What does this mean?

1. One line of code to load it, one line to run it
2. Efficient batching support to generate multiple masks
3. pipeline support for easier usage

More details: 🧵

THIS IS BIG! 👀

It's now possible to take any of the >30,000 ML apps from Spaces and run them locally (or on your own infrastructure) with the new "Run with @Docker" feature. 🔥🐳

See an app you like? Run it yourself in just 2 clicks🤯

Today we are excited to announce a new partnership with @awscloud! 🔥

Together, we will accelerate the availability of open-source machine learning 🤝

Read the post 👉 https://t.co/fb77d1J2qX

It's been an exciting year for 🤗Transformers. We tripled the number of weekly active users over 2022, with over 1M users most weeks now and 300k daily pip installs on average🤯

Hugging Faceから日本へのお知らせです！

Hugging Faceコースの日本語翻訳を始めました。東北大学のStudent Ambassadorsの皆さんのお陰で第一章の翻訳が終了しました。
今後もコツコツと翻訳していきます。
是非コースを読んでHugging Face Tranformersについて学んで、使ってみてください！

Scikit-Learn and 🤗 join forces!

With a growing number of tabular classification & regression checkpoints, we believe statistical ML has its place on the HF Hub.

We're excited to partner with sklearn, statistical ML champion, and move forward together.

https://t.co/j3IVGYjFPv

日本からの嬉しいお知らせです！rinnaが日本語で学習したJapanese Stable DiffusionがHugging Face Spacesでデモ化されました！ https://t.co/Mp8sXdGDV0

Transformers v4.22 is out, and includes the first VIDEO models! 🎥

💥VideoMAE: masked auto-encoders for video
💥X-CLIP: CLIP for video-language

Other nice goodies:
💥Swin Transformer v2
💥Pegasus-X
💥Donut
💥MobileViT

... and MacOS support (device="mps")!

Open Source

🖌️ Stable Diffusion meets 🧨Diffusers!

Releasing diffusers==0.2.2 with full support of @StabilityAI's Stable Diffusion & schedulers 🔥

Google colab:
👉 https://t.co/H3cnpuHVXN

Code snippet 👇

🧨Diffusion models have been powering impressive ML apps, enabling DALL-E or Imagen

Introducing 🤗 diffusers: a modular toolbox for diffusion techniques, with a focus on:

🚄Inference pipelines
⏰Schedulers
🏭Models
📃Training examples

https://t.co/BEboa1yQyE

Last week, @MetaAI introduced NLLB-200: a massive translation model supporting 200 languages.

Models are now available through the Hugging Face Hub, using 🤗Transformers' main branch.

Models on the Hub: https://t.co/DTz4kXcUuR

Learn about NLLB-200: https://t.co/9R5Mh5JPSO

The Technology Behind BLOOM Training🌸

Discover how @BigscienceW used @MSFTResearch DeepSpeed + @nvidia Megatron-LM technologies to train the World's Largest Open Multilingual Language Model (BLOOM):

https://t.co/8QOxhrIVbs

Machine learning demos are increasingly a vital part of releasing a model. Demos allow anyone, not just ML engineers, to try a model, give feedback on predictions, and build trust

That's why we are thrilled to announce @Gradio 3.0: a grounds-up redesign of the Gradio library 🥳

Last week @MetaAI publicly released huge LMs, with up to ☄️30B parameters. Great win for Open-Source🎉

These checkpoints are now in 🤗transformers!
But how to use such big checkpoints?

Introducing Accelerate and
⚡️BIG MODEL INFERENCE⚡️

Load & USE the 30B model in colab (!)⬇️

🤗🚀 https://t.co/M7GGxu5TRa

💫 Perceiver IO by @DeepMind is now available in 🤗 Transformers!

A general purpose deep learning model that works on any modality and combinations thereof
📜text
🖼️ images
🎥 video
🔊 audio
☁️ point clouds
...
Read more in our blog post: https://t.co/5rXpKkgbh2

Transformers v4.13.0 is out and it is *big*:

Vision:
- 🖼️ SegFormer
- 🖨️ ImageGPT

Audio:
- 🔡 Language model support for ASR

Multimodal:
- ⚖️ Vision-Text dual encoders

NLP:
- 🔣 mLUKE
- 🏅 DeBERTa-v3

Trainer:
- 1⃣6⃣ The Trainer now supports BF16/TF32!

🌠New doc frontend 🌠

TODAY'S A BIG DAY

Spaces are now publicly available

Build, host, and share your ML apps on @huggingface in just a few minutes.

There's no limit to what you can build. Be creative, and share what you make with the community.

🙏 @streamlit and @gradio

https://t.co/KyehQt3Z8u

We're thrilled to partner with https://t.co/6jcElMyqbj to create some great new content for their NLP Specialization on Coursera!

With this update, you can access exciting new material and lectures that cover the state of the art in NLP 🧑‍🏫

https://t.co/iMJQxmGvNU

EleutherAI's GPT-J is now in 🤗 Transformers: a 6 billion, autoregressive model with crazy generative capabilities!

It shows impressive results in:
- 🧮Arithmetics
- ⌨️Code writing
- 👀NLU
- 📜Paper writing
- ...

Play with it to see how powerful it is:
https://t.co/v784T0qNOT

🥁 We can't wait to share our new inference product with you! 🤩

- it achieves 1ms latency on Transformer models 🏎
- you can deploy it in your own infrastructure ⚡️
- we call it: 🤗 Infinity 🚀

📅 Join us for a live event and demo on 9/28!
https://t.co/fvhb86gsG7

Document parsing meets 🤗 Transformers!

📄#LayoutLMv2 and #LayoutXLM by @MSFTResearch are now available! 🔥

They're capable of parsing document images (like PDFs) by incorporating text, layout, and visual information, as in the @gradio demo below ⬇️

https://t.co/Hr0kFIPHXW

What’s your go-to local model right now?

1-bit GLM-5.2 GGUF vs. Claude 4.8 Opus vs. GPT-5.5

We gave 3 models the same prompt and compared one-shot outputs.

The 1-bit GLM-5.2 GGUF ran locally on a Mac Studio M3 Ultra with 256GB RAM at ~21.6 tok/s.

Which output do you like best?
GGUF: https://t.co/BMkxswdj5N

You can run GLM-5.2 and other models directly in Unsloth Studio: https://t.co/aZWYAtakBP

GLM-5.2 can now be run locally!🔥

The 2-bit model retains ~82% accuracy after we shrunk it from 1.51TB to 238GB (-84% size).

Run on a 256GB Mac or RAM/VRAM setups.

GLM-5.2 is the strongest open model to date.

Guide: https://t.co/bI7FeeKHDd
GGUF: https://t.co/BMkxswdj5N

You can now run Kimi K2.7 Code locally! 🌘

We shrank the 1T model to 325GB (-48%) via Dynamic 2-bit where important layers are upcasted.

Run at >40 tok/s on 330GB RAM/VRAM setups.

Run full precision on 610 GB.

Guide: https://t.co/SXZJ3IHMpY
GGUF: https://t.co/2lpUx7u0r8

RT @ivanfioravanti: Local AI in action! MiniMax M3 unning locally on a single M3 Ultra 512GB in Unsloth Studio! 🔥

Here UD-Q5_K_XL decoding…

MiniMax M3 can now be run locally!🔥

MiniMax-M3 is a new 428B (23B active) open model with 1M context that performs on par with Gemini 3.1 Pro.

Run Dynamic 2-bit GGUF on 138GB RAM/VRAM or 3-bit on 165GB.

GGUF: https://t.co/lwfWsOBNKl
Guide: https://t.co/EP62nmKK0R

DiffusionGemma can now run at 2000+ tokens/sec! ⚡

We made local DiffusionGemma inference 1.8× faster.

Run it on 18GB RAM via Unsloth Studio.

GitHub: https://t.co/aZWYAtakBP
Guide: https://t.co/wYLfJWE6kG

Gemma 4 now runs 2x faster with MTP GGUFs! Run locally on just 6GB RAM. ⚡️

MTP enables Google Gemma 4 run ~1.4–2.2× faster with no accuracy loss.

Gemma 4 12B MTP can run at 162 t/s vs. 52 t/s without MTP. 31B reaches 101 t/s.

GGUFs + Guide: https://t.co/c4gAUlb6YE

Google releases DiffusionGemma.✨
The new 26B-A4B diffusion text model runs locally on 18GB RAM.

It supports high-speed text generation, thinking, image, video and 256K context.

Run and train via Unsloth Studio.

GGUF: https://t.co/ZH0dCJQ59P
Guide: https://t.co/wYLfJWE6kG

Google releases Gemma 4 QAT. ✨
You can now run Gemma 4 at 3x less memory with near original performance.

Quantization-Aware Training (QAT) makes it possible to run Gemma 4 26B-A4B on 16GB RAM.

GGUFs: https://t.co/wQgEocxUId
QAT Guide: https://t.co/Nsm1yeGEHx

You can now run NVIDIA Nemotron 3 Ultra, a new 550B open model.

Nemotron-3-Ultra-550B-A55B is NVIDIA's largest LLM yet, with 1M context, frontier coding & chat.

Run 2-bit on 200GB RAM, 3-bit on 256GB, 8-bit on 600GB.

GGUF: https://t.co/5WLT7AqUBA
Guide: https://t.co/APXPndWg5V

2-bit Gemma 4 12B GGUF, only 4.66 GB on disk, managed to cite 15 sites from a single prompt.

Try this locally on >6GB RAM via Unsloth Studio.

GitHub: https://t.co/aZWYAtakBP

Vision and audio support for Gemma 4 12B GGUF is now added.

Please update to the latest version of Unsloth and llama.cpp. 🙏

Gemma 4 12B can now run locally on just 8GB RAM via Dynamic GGUFs.

Google's new model, Gemma 4 12B Unified supports image, audio and 256K context.

You can run and train the model via Unsloth Studio.

GGUF: https://t.co/8cL321pVDh
Guide: https://t.co/odRo9WjRpA

Local models are coming to your laptop soon! 🚀

We're excited to partner with @Microsoft to enable millions of developers run local models on Windows!

Introducing Unsloth Studio ✨
A new open-source web UI to train and run LLMs.

• Run models locally on Mac, Windows, Linux
• Train 500+ models 2x faster with 70% less VRAM
• Supports GGUF, vision, audio, embedding models
• Auto-create datasets from PDF, CSV, DOCX
•

Fable 5 is now in Agent Mode - we’ll see if it holds the lead as new traces come in. https://t.co/8ujN06t7FN

Learn more about Agent Arena’s causal tracing methodology with Evan:
https://t.co/e9xKDMXcfC

Check out Peter's first impressions with the model when it first debuted in the Arena: https://t.co/sb4Orq45SF

Fable 5 is back in the Arena!

When it first debuted, Fable 5 ranked #1 in Agent Arena: our benchmark for real-world, long-horizon agentic performance. Agent Arena evaluates models on millions of real tasks submitted by a global community of users, with access to web search,

RT @felicis: What began as a UC Berkeley research project has become one of the fastest-growing and most trusted AI infrastructure companie…

RT @arena: Check out first impressions of Claude Sonnet 5 in the Agent Arena on YouTube with @petergostev.

Scores coming soon.

https://…

Check out first impressions of Claude Sonnet 5 in the Agent Arena on YouTube with @petergostev.

Scores coming soon.

https://t.co/ULnLYdYjFX

See the Video Edit Arena leaderboard details at: https://t.co/d9p8cJDy4K

Gemini Omni Flash ranks #2 for Video Edit in the Video Arena!

With only seven models ranked for this capability, @GoogleDeepMind delivers a strong model (1347) that is nearly +40 points above the next best model: HappyHorse 1.0 (1308)

Congrats to @GoogleDeepMind on the release

Test out all the top agentic frontier models in Agent Mode at: https://t.co/8ujN06t7FN

Learn more about the causal tracing methodology for Agent Arena on our blog: https://t.co/7cotZWljYY

Claude Sonnet 5 is ready for you in the Agent Arena!

In Agent Arena, we measure models on millions of real-world, long-horizon agentic tasks from a global community of users. Models can access web search, filesystem, and terminal tools to complete complex workflows. The

Head over to https://t.co/7NVNbVi1Po to dig into the score details.

Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) ranks #5 overall for Text-to-Image. It has also landed in the Image Edit Arena as:
- #9 Multi-Image Edit
- #15 Single-Image Edit

Gemini 3.1 Flash Lite Image (Nano Banana 2 Lite) has entered the Text-to-Image Arena. It ranks #5 overall.

Google's latest Gemini-3.1-Flash-Lite-Image lands on the Pareto frontier: scoring 1251 at just $0.034/image. This is near flagship quality at a fraction of the price.

The

RT @ml_angelopoulos: So, how does Arena make money?

Our platform serves as a real-world CI/CD system for AI models on post-deployment user…

RT @ml_angelopoulos: Arena has crossed $100M in annualized revenue run rate, eight months after launching our evaluation product.

With our…

If you’d like to learn more about the milestone, head over to our blog: https://t.co/shQ4coO5d4

If our mission resonates with you, check out our job openings and reach out: https://t.co/eLzg6u1tYb

Arena reached a $100M annual revenue run rate just 8 months after launching our evaluation product. We started as a research project at UC Berkeley with a simple mission: measure AI progress through real-world use. As AI shifts from chatbots to agents taking on longer,

Head over to the Video Arena at: https://t.co/ozG8iB7Wry

HappyHorse 1.1 by @HappyHorseATH is in the Video Arena. (Text-to-Video, Image-to-Video & Video Edit)

HappyHorse 1.0 currently holds top 2-4 ranks across the Video Arena, so let's see how the latest version stacks up. Bring your most creative prompts and get voting. Scores coming

Listen in as our engineering team walks through best practices around how to handle billions of data points to keep Arena running: https://t.co/UhLaQlncIE

Millions of people worldwide bring real-world tasks to Arena - and at that scale, hot/cold storage becomes a hard problem fast.

In this clip, our engineering team walks through some best practices: CDC replication, ephemeral storage tradeoffs, and what it takes to build data

Try out Code Arena: Frontend at: https://t.co/DYnDhxb9y7

Introducing Agent Mode: Agentic AI is now measured in the Arena.

Agent Mode can do deep research, create reports, generate images, build websites, debug code, and more.

It completes more complex tasks by using tools like web search, bash in a sandbox environment, image