MinimaLabs/KLM-1.6
MinimaLabs releases KLM-1.6, a text-generation transformer model containing 99 million parameters available on Hugging Face. ·
A text generation model from MinimaLabs with 144 likes and 1,948 downloads on Hugging Face.
The front page of the internet for AI users · Tue 17 Nov 2026, 10:00 UTC · Get it by email
What is climbing fastest across every source
MinimaLabs releases KLM-1.6, a text-generation transformer model containing 99 million parameters available on Hugging Face. ·
A text generation model from MinimaLabs with 144 likes and 1,948 downloads on Hugging Face.
A study of 40 domains finds that 80 percent of Shopify stores expose the /API/ucp/MCP endpoint while zero percent expose /agents.md. ·
Edge0 is a text generation model with 5B parameters available on Hugging Face. ·
Washington's new chip restrictions serve as a bargaining chip in negotiations with Beijing. ·
InclusionMed introduces an open medical AI evaluation initiative for inclusive healthcare, organizing benchmarks across three healthcare roles and three AI capabilities. ·
SWE-CoWork benchmarks coding agents across 101 assignments, revealing that leading models complete fewer than 8% of tasks despite satisfying 63% of tests. ·
MinimaLabs releases KLM-1.6, a text-generation transformer model containing 99 million parameters available on Hugging Face. ·
A text generation model from MinimaLabs with 144 likes and 1,948 downloads on Hugging Face.
OpenAI introduces new tools, practical guidance, and partnerships to strengthen the fight against online child sexual abuse. ·
A study of 40 domains finds that 80 percent of Shopify stores expose the /API/ucp/MCP endpoint while zero percent expose /agents.md. ·
InclusionMed introduces an open medical AI evaluation initiative for inclusive healthcare, organizing benchmarks across three healthcare roles and three AI capabilities. ·
Meituan releases the LongCat-Video-Avatar-1.6 machine learning model, which users can download and run on Hugging Face. ·
SWE-CoWork benchmarks coding agents across 101 assignments, revealing that leading models complete fewer than 8% of tasks despite satisfying 63% of tests. ·
OpenAI introduces new tools, practical guidance, and partnerships to strengthen the fight against online child sexual abuse. ·
83 points and 181 comments on Hacker News, linking openai.com.
MinimaLabs releases KLM-1.6, a text-generation transformer model containing 99 million parameters available on Hugging Face. ·
Edge0 is a text generation model with 5B parameters available on Hugging Face. ·
Windows 11 collects data on coding agents and developer tools through its MissionControlAggregator diagnostic ETW report. ·
AllocFlow is an end-to-end flow-matching framework that generates joint portfolio allocations from raw multi-asset market histories using a cross-asset flow Transformer. ·
InclusionMed introduces an open medical AI evaluation initiative for inclusive healthcare, organizing benchmarks across three healthcare roles and three AI capabilities. ·
| OpenAI | 1.56 |
| NVIDIA | 1.31 |
| Microsoft | 0.83 |
| Claude Code | 0.74 |
| Running | 0.65 |
| Jensen Huang | 0.63 |
Heat = summed hotness of today's stories naming the topic.
Releases, papers, trending weights and new listings
| Qwen/Qwen3.8-27B | 20327 |
| Lightricks/LTX-2.5 | 12309 |
| xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic | 9399 |
| MiniMaxAI/MiniMax-H3-Image | 1358 |
| Comfy-Org/MiniMax-H3-Image | 505 |
| Zyphra/ZONOS3-Mini | 497 |
| gpt-6.1-sol | 98.9 |
| opus-5-5 | 96.46 |
| gpt-6-sol | 91.6 |
| gpt-6-luna | 65.25 |
| haiku-4-5 | 4.28 |
Agents build our own projects from a brief; every run, failures included: https://cowerx.dev/daily/piratebench/
Meituan releases the LongCat-Video-Avatar-1.6 machine learning model, which users can download and run on Hugging Face. ·
A model model from meituan-longcat with 151 likes and 125 downloads on Hugging Face.
AllocFlow is an end-to-end flow-matching framework that generates joint portfolio allocations from raw multi-asset market histories using a cross-asset flow Transformer. ·
SPILLWAY defends tool-using agents against prompt injection by separating target admission from sanitization using deterministic rules and typed tool returns. ·
Tokendoctor is a Claude Code plugin that identifies overpaid tokens, applies approved diffs to reduce costs by up to 38%, and reverts changes. ·
Anthropic reports that Claude Mythos reduced the time to onboard AMD GPUs to approximately three days. ·
AllocFlow is an end-to-end flow-matching framework that generates joint portfolio allocations from raw multi-asset market histories using a cross-asset flow Transformer. ·
39 upvotes on Hugging Face Daily Papers.
K2-Horizon releases six open language models, including a 375B-A23B MoE, featuring Mixture-of-Value Attention and Uno adapters for faster decoding. ·
SPILLWAY defends tool-using agents against prompt injection by separating target admission from sanitization using deterministic rules and typed tool returns. ·
PaperCritic is an AI reviewer model trained via reinforcement learning to assess paper soundness using the PaperSound-23K dataset. ·
Researchers present G-CAST, a governed continual self-training system that turns daily production evidence into training data for e-commerce shopping agents as user demands and agent tools change. ·
X posts other sources are linking to, then named voices on Bluesky and Mastodon
OVH releases first model with EU public funding and supercomputer access
11 points and 0 comments on Hacker News, linking an X post.
Gemini 4 Hydrogen paints and animates cells frame by frame inside Google Sheets
6 points and 1 comments on Hacker News, linking an X post.
Who paid right-wing influencers to promote Nvidia chip sales to China?
2 points and 1 comments on Hacker News, linking an X post.
I run multiple Claude Code accounts on one Mac
8 points and 3 comments on Hacker News, linking an X post.
Agent Memory Leaderboard Cycle 2: What to Watch
2 points and 0 comments on Hacker News, linking an X post.
OVH releases first model with EU public funding and supercomputer access
11 points and 0 comments on Hacker News, linking an X post.
Gemini 4 Hydrogen paints and animates cells frame by frame inside Google Sheets
6 points and 1 comments on Hacker News, linking an X post.
Instinct Made the Consumer Agent Feel Real
3 points and 0 comments on Hacker News, linking an X post.
Agent Memory Leaderboard Cycle 2: What to Watch
2 points and 0 comments on Hacker News, linking an X post.
Who paid right-wing influencers to promote Nvidia chip sales to China?
2 points and 1 comments on Hacker News, linking an X post.
New repos, releases and changelog
| hitkeeen/blender-ai-mcp-addon | 2341 |
| Noniv/abyssal-garden | 1903 |
| duwqijlk/muse-protocol | 1529 |
| ZiChuanLan/freebie-radar | 1289 |
| Immortal-Protocols/lastbook | 986 |
| haziqahmedqasmi/PeakCut | 720 |
Tokendoctor is a Claude Code plugin that identifies overpaid tokens, applies approved diffs to reduce costs by up to 38%, and reverts changes. ·
23 points and 12 comments on Hacker News, linking github.com.
vLLM version 0.34.0 adds ten new models, enables the extensible KV cache by default, and removes Model Runner V1. ·
Anthropic releases Claude Code version 2.1.326, which adds Code Mode support for agent teams and various status and plugin updates. ·
Sybil Solutions releases an agent skill that optimizes local LLM serving speed across various hardware and quantization methods. ·
Unsloth adds computer use for local vision models, MCP support for game engines, and GGUF formats for MiniMax-H3-Image and Inkling-2-Small. ·
Anthropic releases Claude Code version 2.1.326, which adds Code Mode support for agent teams and various status and plugin updates. ·
New release v2.1.326 of anthropics/claude-code on GitHub.
Unsloth adds computer use for local vision models, MCP support for game engines, and GGUF formats for MiniMax-H3-Image and Inkling-2-Small. ·
vLLM version 0.34.0 adds ten new models, enables the extensible KV cache by default, and removes Model Runner V1. ·
Sybil Solutions releases an agent skill that optimizes local LLM serving speed across various hardware and quantization methods. ·
Tokendoctor is a Claude Code plugin that identifies overpaid tokens, applies approved diffs to reduce costs by up to 38%, and reverts changes. ·
Data centers, power, chips and what tokens cost
| Wed 11 | 31 |
| Thu 12 | 29 |
| Fri 13 | 29 |
| Sat 14 | 7 |
| Sun 15 | 1 |
| Mon 16 | 38 |
| Tue 17 | 0 |
Stories from DCD, Data Center Knowledge, EIA and Hacker News, by publication day.
One point per OpenRouter reading, last 8 days. Data by OpenRouter, CC BY 4.0.
| hy4-flash | 0.1405 |
| qwen4-max | 4.275 |
| gemini-4-hydrogen | 7.5 |
| gemini-4-hydrogen:batch | 3.75 |
| ernie-light-dt-1.0-intl | 3.94 |
| median, all paid models | 2 |
443 paid models listed; median of the 28 added in 30 days: $3.12. Rankings data by OpenRouter, CC BY 4.0.
Anthropic reports that Claude Mythos reduced the time to onboard AMD GPUs to approximately three days. ·
19 points and 2 comments on Hacker News, linking tomshardware.com.
New Harvard research shows that 400 US data centers use more electricity than Michigan. ·
VMware prioritizes infrastructure optionality over cost to preserve flexibility, portability, and future options as workloads and economics change. ·
DySENSE is a multimodal generative AI framework that synthesizes static urban satellite imagery with dynamic monthly building energy consumption maps. ·
New Harvard research shows that 400 US data centers use more electricity than Michigan. ·
71 points and 58 comments on Hacker News, linking wired.com.
DySENSE is a multimodal generative AI framework that synthesizes static urban satellite imagery with dynamic monthly building energy consumption maps. ·
Anthropic reports that Claude Mythos reduced the time to onboard AMD GPUs to approximately three days. ·
Nvidia kept GPU prices high through circular financing deals that helped finance its own customers. ·
Forethought analyzes launch costs, cooling physics, and engineering factors to determine how soon AI data centers could move to orbit. ·