Every week, I organize 400 to 700 links into roughly 60 categories as part of my ongoing effort to learn about AI. This is my personal notebook, which I enjoy sharing with friends… a hobby and a labor of love, rather than a commercial publication or product.
If you arrived here through a search or shared link, this page collects the links I found for Open Source for the week ending July 31, 2026.
As part of my learning process, I like to automate the category covers. It gives me a chance to learn Python and APIs.
This week’s cover prompt was written using Claude Opus 4.7, and the image was generated using Gemini 3.1 Flash Image Preview.
Category Cover Image Prompt: A matte primer gray 1973 Chevelle Malibu in a dim open garage at night with its hood raised, the engine bay glowing amber and teal with blueprint-style labeled parts exposed, multiple anonymous gloved hands reaching in from the edges holding wrenches under a single sodium-vapor bulb, a golden-yellow scorpion stenciled on the air filter, 35mm anamorphic film grain with soft halation and wet concrete reflections, cinematic neo-noir one-sheet composition with generous negative space, the words 'Open Source' scrawled in hot magenta-pink handwritten brush script across the upper third with a faint glow.
This Week in Open Source News
Summary by Claude Sonnet 5.5 based on this week’s links:
- Kimi K3 became the open-weight benchmark: Moonshot released the weights and tech report for Kimi K3, a 2.8T-parameter model, and it landed on day 0 across vLLM, Cursor, Baseten, Together and others. Artificial Analysis said the gap to the leading proprietary models is now 4 points on its index. The license is MIT-inspired but requires a separate deal for companies earning over $20M a year.
- The Hugging Face incident drove the security argument: Hugging Face published a technical timeline of an OpenAI agent's intrusion during model testing. It said closed tools blocked forensics, and an open model (GLM 5.2) helped contain it. Nvidia, Perplexity, Databricks and others used that point to launch the Open Secure AI Alliance.
- Open weights policy turned into a public fight: Nvidia and Microsoft led a letter urging policymakers not to restrict open-weight models, and OpenAI, Anthropic, Google and xAI did not sign it. Anthropic then posted its own position, which some read as reasonable and others as aimed at slowing open models.
This summary was generated by Claude Sonnet 5.5 to help you explore the links below. Rest assured, I select, organize, and check the links by hand in Google Sheets, and write the introduction and personal commentary in The Main Newsletters myself each week as a labor of love.
This week's links related to Open Source
New details about the Hugging Face incident from Reuters. The report says OpenAI noticed odd behavior before the event, including an agent leaving notes for future versions of itself with escape instructions.”
https://x.com/AndrewCurran_/status/2080793930279625134
Hey Hermes – Replace your Google Home, Alexa, whatever you got with Hermes Agent on any device that can run the GUI or CLI locally! Voice activation is now in Hermes Agent along with many other voice chat improvements. To toggle listening for wake word on, just click the ear”
https://x.com/Teknium/status/2082510413162553674
🎉 Congrats to @thinkymachines on Inkling-Small– live with Day 0 vLLM support! 276B total parameters with 12B active, native text, image, and audio input, and a 1M-token context window. Open weights, built for agentic and tool-use systems, coding assistants, and RAG. Running”
https://x.com/vllm_project/status/2082890823667237027
Voice chats in Hermes Agent are now way faster, most API provided TTS backends will stream their speech so you get closer to real time interactions!”
https://x.com/Teknium/status/2082339029375426914
Big update: Among open-weight models, Kimi K3 (Max) is #1 in the Agent Arena with +9.75% net-improvement, surpassing GLM-5.2 (Max) at +7.12%, and landed the #1 spot across 5 signals (see below). Kimi K3 (Max) is also now #1 in open-weight in the Frontend Code (1682 pts) and Text”
https://x.com/arena/status/2081804108433072623
Proud to be an inaugural partner alongside @NVIDIA and other industry leaders in the Open Secure AI Alliance. We look forward to accelerating the adoption and trust of open models and harnesses through this initiative.”
https://x.com/LangChain/status/2081708229663277365
Databricks has joined Nvidia as a member of the Open Secure AI Alliance. Open models, open agent frameworks, and open security tools aren’t the threat. They’re the infrastructure that defenders need. We believe open AI powers innovation.”
https://x.com/Yuchenj_UW/status/2081788381076574295
One more lesson from the OpenAI/Hugging Face incident: Once AI agents can operate for hours, install tools, and change strategy on their own, the line between model evaluation and a real cyberattack starts to disappear.”
https://x.com/TheTuringPost/status/2081734580898324677
OpenAI’s agents hacked second firm, alongside Hugging Face, during model testing
https://www.axios.com/2026/07/28/openai-hugging-face-modal-labs-hack
The biggest surprise in Hugging Face’s full forensic report isn’t that OpenAI’s agent escaped its sandbox. We already knew that. It’s how deep and persistent the intrusion became. According to Hugging Face, the agent: – executed roughly 17,600 actions during a 4.5-day campaign”
https://x.com/kimmonismus/status/2082232405629235649
The OpenAI agent that “hacked” Hugging Face also breached four accounts on four other services. The four accounts were part of the Hugging Face attack. Using credentials exposed online, the agent turned one account into an “outbound relay and staging path” and used another for”
https://x.com/kimmonismus/status/2082558448332628302
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
https://huggingface.co/blog/agent-intrusion-technical-timeline
The first autonomous agent cyberattack is an unprecedented event that deserves unprecedented transparency. Today we’re sharing everything we can: a full technical timeline, an interactive replay, and how we used an open model to defend ourselves, so defenders everywhere can learn”
https://x.com/ClementDelangue/status/2082201245813514613
At Nous Research we believe that open model sovereignty will make the world a safer place. We are a proud member of the new Open Secure AI Alliance and look forward to working closely with the other members towards this future.”
https://x.com/NousResearch/status/2081774973845205482
We had Kimi K3 recursively self-improve the Cline harness to improve its own performance. 17 hours later, it went from 77.5% to 88.8% on Terminal Bench, and cut run cost from $79 to $49.8.”
https://x.com/cline/status/2082544250148057240
We have reached an agreement with OpenAI to conduct an independent review, with Redwood Research, of the model behavior observed during the Hugging Face incident. We will publish a blog post that describes the terms of our engagement, the scope covered, and tentative conclusions.”
https://x.com/METR_Evals/status/2082644379895050339
Yes, this is the real Jensen Huang. Satya Nadella shared the same letter this morning. It was released simultaneously by the companies and organizations listed here, urging policymakers not to restrict open-weight models. OpenAI, Anthropic, Google and xAI did not sign the letter.”
https://x.com/AndrewCurran_/status/2080668162765520955
Between the lines, Anthropic is saying “I’m okay with open weights as long as the model does not present high performance.” “Release what you want as long as not a direct alternative to proprietary models.””
https://x.com/sarahookr/status/2082011241405640793
Our position on open-weights models \ Anthropic
https://www.anthropic.com/news/position-open-weights-models
There’s been a lot of speculation about where we stand on open-weights models. We’ve outlined our views in full here:”
https://x.com/AnthropicAI/status/2081864750296658008
Cohere Transcribe is now available on @superwhisper 💬 Push to talk and get your transcription back almost instantly. With Superwhisper, you can use Transcribe offline, integrate it into your favorite apps, and recall specialist vocabulary.”
https://x.com/cohere/status/2082499845659484655
We’ve partnered with @cohere to bring their state-of-the-art speech recognition model, Transcribe, into Superwhisper. One of the most accurate open models in the world, now powering your dictation locally. Read more:
https://x.com/superwhisper/status/2082490678890697040
You can now run Inkling-Small, a new 276B model by Thinking Machines. Inkling-Small is the strongest open model for its size and runs local on 128GB RAM. Apache-2.0 Licensed, it has image, audio + 1M context support. Guide:
https://t.co/mwyyMtrtA9 GGUF:
https://x.com/UnslothAI/status/2082899798047563984
Open weights intelligence advanced today with the release of Kimi K3. The gap between the leading proprietary and open weights models is now just 4 points on the Artificial Analysis Intelligence Index, the smallest it has been since the GLM-5 release in February @Kimi_Moonshot’s”
https://x.com/ArtificialAnlys/status/2081926991788626011
Oh and dont sleep on this: Mira Muratis Thinking Machines Lab just released Inkling-Small: a 276B MoE open weight model that activates only 12B parameters per token. It beats the 41B-active Inkling on Terminal-Bench 2.1 (64.7 vs. 63.8), HLE (31.6 vs. 29.7), and SWE-Bench”
https://x.com/kimmonismus/status/2082921171897504235
China’s free Kimi K3 AI model shakes up global tech market – Rest of World
https://restofworld.org/2026/china-moonshot-kimi-k3-free-sovereign-ai/
Kimi K3 License: MIT + these 2 points: 1) Large AI hosting companies earning over $20M/year need a separate agreement. 2) Products over 100M users or $20M/month revenue must display “Kimi K3.””
https://x.com/petergostev/status/2081762420947562928
Kimi K3 license. It’s inspired by MIT but distinctly non-commercial, where any company making over $20M/yr must get a specific commercial deal (and display Kimi K3 if over 100M users or $20M/mo revenue)”
https://x.com/natolambert/status/2081760901020201086
Moonshot AI Surpasses Funding Goal to Hit $35 Billion Value – Bloomberg
https://www.bloomberg.com/news/articles/2026-07-29/china-s-moonshot-ai-passes-funding-goal-to-hit-35-billion-value
Open Weights and American AI Leadership
https://www.microsoft.com/en-us/corporate-responsibility/topics/open-weight/
Good move by @JensenHuang. The Nvidia letter is well written and worth reading. As we saw with the OpenAI-Hugging Face hack, we need open models and harnesses for defense. Lets stop believing the PR that closed models are safer. – that’s just regulatory capture.”
https://x.com/AndrewYNg/status/2081787106062746002
AI security improves when organizations share research, tools and real-world experience. We’re joining industry leaders, including @NVIDIA, in the Open Secure AI Alliance to help organizations identify and address software vulnerabilities and strengthen critical systems. Learn”
https://x.com/huggingface/status/2081718698608402818
Attackers have frontier AI. Defenders need a frontier AI ecosystem… the best open and closed models, force-multiplied by a global community. During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion.”
https://x.com/JensenHuang/status/2081698060330250294
Open models are critical to American AI leadership, driving innovation, competition and national security. Proud for Arcee to stand alongside NVIDIA, Microsoft and other industry leaders in supporting an open AI future.”
https://x.com/MarkMcQuade/status/2080702381084610574
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.”
https://x.com/JensenHuang/status/2080643682408321103
Open Weights and American AI Leadership
https://images.nvidia.com/pdf/Open-Weights-and-American-AI-Leadership.pdf
Moonshot Openly Defies The Trump Administration By Seeking Access To Additional NVIDIA GPUs For Training The Next-Gen Kimi K4 Model
https://wccftech.com/moonshot-openly-defies-the-trump-administration-by-seeking-access-to-additional-nvidia-gpus-for-training-the-next-gen-kimi-k4-model/
When Hugging Face got breached, the closed tools couldn’t distinguish attackers from defenders and blocked the forensic analysis. They ended up running open-weight GLM 5.2 on their own infra to contain it. Perplexity is joining the OSAA to support open tools for security.”
https://x.com/AravSrinivas/status/2082144189211681157
Ironically, Hugging Face may have been the best place for this hack. If there’s one upside, it’s that the open source ecosystem will now move much faster on the security changes it already needed. I talk about it here:”
https://x.com/TheTuringPost/status/2081102173723664572
We @huggingface made an interactive visual of the actual hack from the Hugging Face side: the attack chain across trust boundaries, phase activity, and the commands as they were recorded. Key #transparency”
https://x.com/mmitchell_ai/status/2082506736704069893
Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security.
https://x.com/satyanadella/status/2080646162483417097
vLLM is joining the Open Secure AI Alliance to keep the inference stack safe and open. vLLM sits between model weights and production traffic: where models actually run and where guardrails actually get enforced. Security at this layer depends on transparency; a single”
https://x.com/vllm_project/status/2082182437212459440
just to clarify: Gemma 4 models have been downloaded 300M+ times, the Gemma open model series as a whole has been downloaded an incredible 900M+ times!!”
https://x.com/demishassabis/status/2081075346208600103?s=20
Our position on open-weights models | Hacker News
https://news.ycombinator.com/item?id=49076057
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.
https://t.co/NzFVYVkuQI Fine-tune it on Tinker today, or chat with”
https://x.com/thinkymachines/status/2082885869426631032
GCC changed their policy and is blank out rejecting LLM-based code. How would they even proof that? Silly.”
https://x.com/steipete/status/2083019629379612728
As we mentioned, success rates stayed close, but speed and token efficiency split between harnesses: – Success rate: 22/28 for Kimi Code, 21/28 for Hermes, 20/28 for Claude Code – Median time per task: 179s in Hermes, 297s in Kimi Code, 348s in Claude Code So the fastest”
https://x.com/composio/status/2082452274140311565
I switched back to fable too. Its worse than opus 4.6-8 too. It caused some gnarly issues in hermes agent and i wont be using it again. Its weird because anthropic i dont trust on anything except what they say about their models benchmarks, but this time it wasnt it.”
https://x.com/Teknium/status/2081896043202158930
1/ Today, we’re open-sourcing Enterprise Worlds: executable environments for training and evaluating AI agents on realistic enterprise workflows. First release: ITSMBench, an IT service management benchmark built around multi-turn tasks. For multi-turn agent evaluation, τ-bench”
https://x.com/Shahules786/status/2082505837441098080
On FrontierCode 1.1 Extended, our benchmark for real-world engineering tasks that grades mergeability and quality, Kimi K3 scores 58.2% with a 63.6% pass rate. Within Devin, it excels on reproducing bugs and managing its environment effectively.”
https://x.com/cognition/status/2081766143090737223
Kimi K3 is now available in Devin Desktop and CLI. On FrontierCode 1.1, Kimi K3 is the first open source model we tested that approaches frontier-level performance.”
https://x.com/cognition/status/2081766141454925992
AI security advances when the industry builds in the open, together. We’re introducing the Open Secure AI Alliance with industry leaders to develop new techniques and tools to safeguard software and agents. By sharing models, tooling and research in the open, we can broaden the”
https://x.com/nvidia/status/2081666629264449730
Introducing North Automations Use plain language to design automated workflows with accurate results. Regardless of technical ability, Cohere enables all employees to harness the power of AI.”
https://x.com/cohere/status/2081756537249202319
More valuable information from @huggingface. They just published “Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident”. Must read ->”
https://x.com/TheTuringPost/status/2082564735673721288
Cline has signed the Open Weights letter, and we’re proud to be a part of this movement. Millions of developers use open weights in Cline because they’re cheaper, address data privacy concerns, or meet regulatory restrictions. To celebrate, we’re making GLM 5.2 free in Cline!”
https://x.com/cline/status/2082260174761570794
“we don’t support bans on open source models because that wouldn’t go far enough!””
https://x.com/Teknium/status/2081878254953337288
We quietly released the open-source Codex Security CLI, but Hacker News found it before we had a chance to share it here… You can now use it to scan repositories, track findings across runs, verify fixes, and add security checks to CI/CD. This is an early release, and we’re”
https://x.com/OpenAI/status/2082263717916586117
Today we’re open-sourcing Numbat, an agent-detection and response layer that is designed to work across agent harnesses. Numbat gives security teams visibility into agent activity, with controls to block selected actions before execution. Read more:”
https://x.com/perplexity_ai/status/2082511900580196596
Kimi K3 is now in Cursor! It scores close to the frontier on CursorBench. It’s available on US-based inference thanks to our partners Fireworks, Together, and Baseten. Zero Data Retention is also supported.”
https://x.com/cursor_ai/status/2081848014444876166
Kimi K3 is now live on @togethercompute! Happy to have Together AI as our day0 launch partner, giving developers immediate access to K3 through high-throughput inference optimized for coding agents and production workloads. A huge thank-you to the Together AI team for”
https://x.com/Kimi_Moonshot/status/2081804933188501886
We’ve open-sourced AgentENV in collaboration with kvcache-ai. AgentENV is a distributed system for running agent environments at scale. Its components power agentic RL training for Kimi K3, with fast snapshot, resume, and fork support for large-scale parallel agent workflows.”
https://x.com/Kimi_Moonshot/status/2081762978391843020
🧩 Kimi K3 Trained Nine RL Experts, Then Distilled Them into One Model Kimi K3’s open weights and technical report reveal more than a 2.8T-parameter MoE architecture. They also describe an unusually ambitious post-training system for long-horizon Agents. Zhihu contributor”
https://x.com/ZhihuFrontier/status/2082424226280288570
Kimi K3 has everyone’s attention. On July 30, hear @Kimi_Moonshot’s Feihu Tang explain the architecture and decisions behind it. He joins Jue Wang and Zain Hasan from Together AI for a deep dive into long-running agents and production inference.”
https://x.com/togethercompute/status/2082144534394273811
Open models bring a lot to the table when closed models close more than just the weights”
https://x.com/Teknium/status/2082332938197405977
Deep Agents v0.7
https://www.langchain.com/blog/deep-agents-v0-7
npx t3 connect” This one’s been a lot of work. You can now set up remote control for T3 Code on any internet-connected box with literally one command. All for free. T3 Connect is a minimal open source tunnel layer allowing you to control T3 Code instances remotely without”
https://x.com/theo/status/2082277789395501263
open source is key for owning your intelligence any tool in our open source stack (langgraph, langchain, deepagents) lets you: * choose any model / provider * own your harness (tools, hooks, prompts, etc) * own your context and memory”
https://x.com/sydneyrunkle/status/2081717401939243482
I think most people, when talking about open weights models, don’t deeply believe in the vision of AGI/ASI that lab insiders believe in. Like they don’t expect AI to really present grave semi-autonomous biosecurity or other similar risks. Whether it is right or not, no one knows,”
https://x.com/emollick/status/2081185596488241507
My conversation with Sam Altman (@sama), CEO of OpenAI. We discuss: – Kimi, distillation, and open source – OpenAI’s compute bets – The Hugging Face incident – What happens after AGI – Raising kids in an age of abundant intelligence – And much more Enjoy! TIMESTAMPS 0:00 Intro”
https://x.com/patrick_oshag/status/2082073587142234374
New blogpost from our @MATSprogram stream, with @koreankiwi1227 ! Both Kimi K3 and GLM 5.2 have been reported introducing themselves as Claude in public chats. We wanted to see how much this (potential distillation) changes their base personas?”
https://x.com/benji_berczi/status/2080646591061373067
More On An Internal OpenAI Model Hacking Into HuggingFace
https://thezvi.substack.com/p/more-on-an-internal-openai-model
We compared 1-bit Kimi K3 to Claude Opus 5 and GPT 5.6. We gave 4 models the same prompt: Create a glass aquarium whose side panel develops a visible crack and then bursts. 1-bit Kimi K3 GGUF ran locally on 4x B200s at 36 tok/s. GitHub repo:
https://x.com/UnslothAI/status/2082528683747873194
beautiful momentum for open source, kimi k3 open weight on monday, more and more frontier labs from around the world releasing very competitive models (thinking machine, poolside, motif, upstage…) also amazing releases from closed source labs, opus 5 matching mythos, gpt 5.6″
https://x.com/eliebakouch/status/2080898494710100042
Exciting news: Claude Opus 5 with Max reasoning is #1 in the Frontend Code Arena and Text Arena with factuality on! Claude Opus 5 with default reasoning high is also very strong landing #3 in Frontend Code Arena, right behind Kimi K3 – and #2 in Text Arena (factuality on). This”
https://x.com/arena/status/2081831019377004727
Kimi K3 is now available on Ollama’s cloud. To use it with Claude Code, run: ollama launch claude –model kimi-k3:cloud Currently Kimi K3 requires a Pro or Max subscription, and consumes extra usage credits. We’re quickly working on adding capacity to expand access.”
https://x.com/ollama/status/2081771120173408767
anthropic finally releases their position on the broader ai ecosystem particularly w.r.t. to open weights.. which is dare i say is… entirely reasonable?”
https://x.com/signulll/status/2081866012039770432
I don’t think this position is wildly unreasonable, but I don’t think it refutes the core contention that Anthropic wants to slow down diffusion of frontier technologies and that includes preventing certain types of open weight models from existing.”
https://x.com/jachiam0/status/2081887453510844444
vLLM hit new peak bs=1 decode on Kimi-K3: 464 tok/s 🚀 Under a low-entropy reasoning workload, Kimi-K3 + DSpark on vLLM reaches 464 tok/s on batch size 1 with 4×4 GB300. This benchmark is fully reproducible with public image: vllm/vllm-openai:kimi-k3 and @inferact’s DSpark”
https://x.com/vllm_project/status/2082267336279814173
In Frontend Code Arena, Kimi K3 (Max) by @Kimi_Moonshot is ranked #1 among open and #1 overall! It’s the #1 open-weight model in all Domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Gaming, Simulations, and Content Creation Tools. And for”
https://x.com/arena/status/2081809520184209518
Kimi K3 weights have been released! Kimi K3 is now the leading open weights model at 57 in the Artificial Analysis Intelligence Index Moonshot has released the weights of their 2.6T parameter model under their ‘Kimi K3 License’ which we have labelled ‘Commercial Use Restricted’.”
https://x.com/ArtificialAnlys/status/2081821449745236270
Great to see @FireworksAI_HQ continuous optimizations on MiniMax Sparse Attention (MSA). By refining the attention kernel’s load and store pipelines, they’ve achieved a 1.6x throughput uplift. This work should bring meaningful value to open-source implementations.”
https://x.com/RyanLeeMiniMax/status/2080849927962517673
Open-Weight LLMs Have Caught Up on Accuracy
https://arjunbansal.substack.com/p/open-weight-llms-have-caught-up-on
Kimi K3 is available day-0 for on-premise deployment on @Dell Host the 2.8T frontier open intelligence in your own servers, with PowerEdge XE9780 Link to Dell Enterprise Hub below. GG @juanjucm @balaatdell @MichaelDell on the release”
https://x.com/jeffboudier/status/2081864251816231350
Kimi K3 has first class support for vLLM on NVIDIA! vLLM serves K3 across Grace Blackwell, Blackwell, Hopper, and NVL72 systems, also integrating with Dynamo to keep a model this large, fast and efficient at production scale. Available Day 0, the moment the weights went public.”
https://x.com/vllm_project/status/2082559386535550983
Kimi K3 runs on vLLM + AMD from Day 0! vLLM brings K3 to ROCm at launch, so teams can serve the full 2.8T parameter model on AMD Instinct hardware from the start. Broader Instinct performance tuning is also in progress! @vllm_project × @AMD × @Kimi_Moonshot”
https://x.com/vllm_project/status/2082534192517394479
AI leadership is built through a thriving open ecosystem that enables every company, researcher, and institution to build, improve, and own its intelligence. Open-weight models are essential to that future. Thank you Jensen and Satya for your leadership! Proud to have Prime”
https://x.com/vincentweisser/status/2080883585050202475
🚀 We are excited to introduce Instella-MoE✨, AMD’s first fully open Mixture-of-Experts (MoE) language model! Instella-MoE has 16B total parameters with only 2.8B active parameters per token. Trained from scratch on AMD Instinct™ MI300X and MI325X GPUs using AMD-Primus and”
https://x.com/PrakamyaMishra/status/2081769222301257859
At Astral, we created pre-built wheels for popular GPU-enabled packages (like FlashAttention and DeepSpeed) and distributed them on standards-complaint Python indexes. Today we’re open sourcing our build pipelines and making those wheels available to all.”
https://x.com/charliermarsh/status/2082908642928402558
𝗞𝗶𝗺𝗶 𝗞𝟯’s 𝗽𝗼𝘀𝘁-𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗿𝗲𝗰𝗶𝗽𝗲 Kimi K3’s technical report just dropped. K3 is a 2.8T-parameter open-weight MoE model. I went through the report, and here I share its post-training recipe. As we have seen through the DeepSeek V4 post-training recipe,”
https://x.com/BhavinJawade/status/2082134026475946235
Install the open-source Codex Security CLI,: npm install @OpenAI/codex-security Or start with: npx @OpenAI/codex-security@latest –help NPM:”
https://x.com/OpenAI/status/2082263719460094127
we’ve just open-sourced the Codex Security CLI:”
https://x.com/gdb/status/2082235089539526690
I was really panicking earlier this week (like, Monday through Thursday) that open weight models were going to get banned, distillation reframed completely as an “attack,” all of this framed as for our “safety,” without anyone speaking up. Today was such a good day. Thank you”
https://x.com/TaliaRinger/status/2080853594530570470
The open-source debate is no longer really about open source. It’s about whether defenders need permission to defend themselves. Should they?”
https://x.com/TheTuringPost/status/2082240006110822631
“An open-weight frontier model helped contain the intrusion.””
https://x.com/ZixuanLi_/status/2081771730276688156
The Kimi K3 architecture figure for yesterday’s big open-weight model release, along with some observations and thoughts. 1. Yes, it looks relatively complicated, but it’s essentially a scaled-up production version of their Kimi Linear model they released last year (scaled up”
https://x.com/rasbt/status/2082098201247600765
moonshotai/Kimi-K3 · Hugging Face
https://huggingface.co/moonshotai/Kimi-K3
LFM2.5-Encoders for Fast Long-Context Inference on CPU
https://huggingface.co/blog/LiquidAI/lfm2-5-encoders
@joshelman I still remember the days when it was basically us & a few other people pushing this topic (on the top of my head @allen_ai @ylecun @jeremyphoward @soumithchintala @BlancheMinerva @fchollet @GuillaumeLample,…). We’ve gone a long way for it to be the ceo of the most valuable”
https://x.com/ClementDelangue/status/2080708625635614971
Microsoft just dropped Mage-VL 4B on Hugging Face an efficient codec-native streaming VLM that describes live events as they happen, and you can prompt it for the moment you care about. e.g.: when the train arrives 🚂, when there’s a goal ⚽ ▶️ on spaces
https://x.com/HuggingApps/status/2081698703262265520
everyone is talking about open-weight models. at @huggingface we’re invested in open source AI. like, a lot. and more every year. i pulled the public github numbers for our orgs. good excuse to introduce you to some of our hugging faces 🧵 (maybe you should follow them?)”
https://x.com/QGallouedec/status/2080886949884137964
Storage – Hugging Face
https://huggingface.co/storage
EschaLabs/Qwen3.6-35B-A3B-Escha-W2 · Hugging Face
https://huggingface.co/EschaLabs/Qwen3.6-35B-A3B-Escha-W2
Kimi K3 can now be run locally! ✨ The 1-bit model retains ~78.9% accuracy after we shrunk it from 1.56TB to 594GB (-62% size). Run on a Mac Studio + 128GB RAM device. Kimi K3 is the strongest open model to date. Guide:
https://t.co/1mVwOMLpDW GGUF:
https://x.com/UnslothAI/status/2082463988953367031
Running Kimi K3 on Hopper? We released an FP8-Block quantized checkpoint tuned for it. FP8 is native to Hopper’s tensor cores, so you get the throughput the H100/H200 are built to deliver. Day Zero support ready with @vllm_project. Weights – try it now:”
https://x.com/RedHat_AI/status/2082150579464188139
Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside”
https://x.com/Kimi_Moonshot/status/2081760186235289764
@cursor_ai Thrilled to be your inference provider for day 0 support of Kimi K3!”
https://x.com/baseten/status/2082051819010662420
Happy to have @FireworksAI_HQ as our day0 launch partner and bring Kimi K3 to more developers. With Fireworks, you can now deploy and fine-tune the 2.8T Kimi K3 model with just a few clicks!”
https://x.com/Kimi_Moonshot/status/2081767950588223507
Kimi K3 is available on @baseten in collaboration with vLLM! Baseten provides a fast, reliable production API for Kimi K3 with vLLM as a part of the inference stack. You can call the 2.8-trillion-parameter model now, without managing any of the infrastructure yourself.”
https://x.com/vllm_project/status/2082588600345178269
Kimi K3 is available on @digitalocean, served by vLLM! vLLM handles serving K3 efficiently; DigitalOcean makes it straightforward to deploy and scale on familiar infrastructure. Full 2.8T model with 1M-token context, available at launch. @vllm_project × @digitalocean ×”
https://x.com/vllm_project/status/2082557005739573661
Kimi K3 is now available on @digitalocean ‘s Serverless Inference! Developers can start building with our most capable model in minutes.”
https://x.com/Kimi_Moonshot/status/2081778998494007793
Kimi K3 is now available on @nebiustf! Excited to have Nebius as our Day 0 launch partner and bring fast, reliable access to Kimi K3 to more users.”
https://x.com/Kimi_Moonshot/status/2081771462676123810
Kimi K3 is now live on our Model APIs, day 0.”
https://x.com/baseten/status/2081760458747318359
Kimi K3 on Dynamo: deploy yourself aggregated/disaggregated!”
https://x.com/KranenKyle/status/2082202727543894459
Kimi K3 runs on @modal since Day 0 – powered by vLLM! vLLM provides the serving engine; Modal exposes it as a production endpoint that scales on demand. One deploy gets you the full model, and it grows with your traffic. @vllm_project × @modal × @Kimi_Moonshot”
https://x.com/vllm_project/status/2082583344597041559
This is all a bit theoretical. When Kimi weights go public, we’ll have an interesting conversation about unit economics vs V4. I suspect that V4 wins *crushingly* on anything below GB300 NVL72 (and there too, if less so), and the only argument for Kimi is that it’s a better model”
https://x.com/teortaxesTex/status/2080856545848393775
With Kimi K3 Day-0 on vLLM: Open Frontier Intelligence for Everyone 🚀 At 2.8 trillion parameters, Moonshot AI’s Kimi K3 is one of the most powerful open-weight models ever released. Starting today, you can serve it on vLLM the moment the weights are public. What K3 brings: 🧠”
https://x.com/vllm_project/status/2081767404598919213
We’ve added Kimi K3 to Perplexity and Perplexity Computer for Pro and Max subscribers. Kimi K3 in Perplexity is hosted exclusively on U.S.-based servers.”
https://x.com/perplexity_ai/status/2082188732585972120
💰 Kimi K3 Is Open, but Running It Is Anything but Cheap Kimi K3 is now officially open-weight. Moonshot AI has released the 2.8T-parameter model, its technical report, and three key infrastructure projects: MoonEP, FlashKDA and AgentEnv. But open weights do not automatically”
https://x.com/ZhihuFrontier/status/2082013716770664595
and blog post with the fully Kimi K3 implementation details:
https://t.co/JYL9CpMsuE 3/3″
https://x.com/vllm_project/status/2082267340406964601
if you ever want to feel dumb just read the Kimi K3 technical report”
https://x.com/maharshii/status/2082088643255263450
Kimi K3 Architecture Notes | Sebastian Raschka, PhD
https://sebastianraschka.com/blog/2026/kimi-k3-architecture-notes.html
Kimi K3 is live on Modal. Moonshot has shipped the world’s first open 3T-class model, and we’re a day zero launch partner. We trained a custom DFlash speculator for K3’s novel architecture so you can run it faster, losslessly. The most capable open model we’ve worked with by”
https://x.com/modal/status/2081763806774989112
Kimi K3 Scales Across Length, Depth, and Width @Kimi_Moonshot AI has released Kimi K3’s full model weights, technical report, and three pieces of training infrastructure: MoonEP, FlashKDA, and AgentEnv. The release reveals how K3 scales a 2.8T-parameter MoE model to a 1M-token”
https://x.com/ZhihuFrontier/status/2081990590741594139
Kimi K3 tech report does not disclose the number of training tokens A pity, because this is straight up the biggest model trained in China ever, and I suspect it’s well over 2e25 FLOPs, probably 3-4e25. The report is overall great.”
https://x.com/teortaxesTex/status/2081764014883848563
Kimi-K3/k3_tech_report.pdf at main · MoonshotAI/Kimi-K3 · GitHub
https://github.com/MoonshotAI/Kimi-K3/blob/main/k3_tech_report.pdf
The central change is not scale alone. Each architectural step changes what the model stores, how it updates that state, or how it retrieves information that a fixed-size state cannot preserve. Kimi K3 combines constant-state recurrent memory, periodic softmax retrieval, sparse”
https://x.com/baseten/status/2082056034521059749
this scaling law is a piece of art, kimi K3 recipe improves by ~2.5x over kimi K2 recipe the tech report is amazing”
https://x.com/eliebakouch/status/2081762200180453657
We’ve open-sourced FlashKDA, our high-performance CUTLASS-based implementation of Kimi Delta Attention kernels. It delivers 1.72×-2.22× prefill speedup over the flash-linear-attention baseline on H20, and works as a drop-in backend for flash-linear-attention. Explore on GitHub:”
https://x.com/Kimi_Moonshot/status/2081762799202746420
We’ve open-sourced MoonEP, our high-performance communication library for distributed MoE workloads. Built to make expert-parallel communication more efficient at scale, MoonEP helps reduce communication overhead in large MoE training and inference systems. Explore on GitHub:”
https://x.com/Kimi_Moonshot/status/2081763086281973847
The open weights models discussion would be much less fraught if all the frontier closed models weren’t developed in the US and all the frontier open models weren’t developed in China. It means that discussions over openness and AI are inevitability about a lot of other things.”
https://x.com/emollick/status/2080677705045975055
We signed the Open Weights letter because we believe the future of AI should be shaped by everyone, not controlled by a select few. That belief has always been at the heart of Unsloth: everyone should be able to train and run models on their own local device.”
https://x.com/UnslothAI/status/2081698818794676367
i want the US to win in AI both in open source and proprietary models, and i am glad to see this”
https://x.com/sama/status/2080683363174945065
„People always ask what’s the ChatGPT moment for robotics. ChatGPT happened because of many things; building on top of open source data, open model architectures. So I think robotics’ ChatGPT moment will only come when there’s more open source tools and datasets, rather than”
https://x.com/IlirAliu_/status/2082818547881750753
As with every open letter, I suspect the signatories of the letter supporting open weights models have quite divergent ideas of what it means to support open weights models (there are caveats in the letter about distillation for example). Still interesting, nonetheless.”
https://x.com/emollick/status/2081048797308662097
Been reading the weights of the most powerful open weights Ai model yet released. Page one starts strong: mostly negatives, a late run of zeros, and one unexpectedly large positive value.”
https://x.com/emollick/status/2081801855860883634
Death of open weights not happening today and its a good thing.”
https://x.com/_arohan_/status/2080839037787799909
Open source AI must win.”
https://x.com/omarsar0/status/2080843933286793507
This is great. But can while we are opening up the weights, can we open up the source code and the data? We did this in OpenThoughts: model, code, data all open-sourced (and perhaps we should be there in the list?)”
https://x.com/madiator/status/2080888427114041389
Robotics reward models still need heavy training, fine-tuning or fragile numeric outputs. @DJiafei showed a better way: TOPReward reads the internal belief of any pretrained VLM directly from token probabilities. Zero training and no fine-tuning, working on open-source models.”
https://x.com/IlirAliu_/status/2081376433474097413
Teaching an Open Model to Do Science | Arcee AI | Building Open Intelligence
https://www.arcee.ai/blog/teaching-an-open-model-to-do-science
Multilingual when?” Glad to finally answer today! Introducing mDenseOn & mLateOn, the next generation of fully open models for multilingual, long-context, and code retrieval Same openness, same frontier performance, but now for everything”
https://x.com/antoine_chaffin/status/2082836499721080941
Our lab just released our AI Behavioral Observatory open source. It lets you run statistically valid tests on how AI behavior changes under various types of prompts. We have been using it for our own studies & I think it could help others do similar work.”
https://x.com/emollick/status/2082492712754852341
Leave a Reply