☀ TL;DR

DeepSeek open-sources Ultrafast inference engine with 1M-token context

DeepSeek open-sourced its Ultrafast inference engine for DeepSeek-V4.1-Flash, supporting 1M-token context and 2,060 tokens/s decode on eight B300 GPUs.

Anthropic added coordinator messages, thread sharing and /fork branching to Claude Code Projects in a public beta for Pro and Max.

The U.S. Treasury Secretary questioned whether Nvidia knowingly sold AI GPUs to shell companies.

A law professor said Claude deleted a syllabus and a year of notes while renumbering pages.

Industry

24
24h Key Points
  1. Gemini 4 Carbon · Along with Anthropic's Fable 5.2, knocks GPT-6.1 Sol off the MathArena top spot, taking first place at roughly 15 times its cost and almost never giving up on tasks.

  2. Anthropic · Makes Claude Opus 5.5 the default model on all paid plans and publishes a prompting guide telling users to drop old habits like "think hard" and start at Medium effort.

  3. US Treasury · Treasury Secretary questions whether Nvidia knowingly sold AI GPUs to shell companies and says individuals at the firm should be held accountable; possible export-control or sanctions issues.

  4. Helen Toner · The former OpenAI board member who voted to fire Sam Altman says trusting any single person is a serious risk and that the nonprofit board's accountability mechanism failed.

  5. GLM-5.3 · The 753B-parameter MoE model, quantized to NVFP4 with ~464GB of weights, runs on two DGX Stations via vLLM at TP=2 and reaches 607.8 tok/s.

  6. Multiverse Computing · With the University of Toronto, demonstrates a multi-head quantum neural network replacing transformer blocks in a 30B-parameter LLM, with 145,281 trainable parameters displacing 16.33% of model parameters while keeping accuracy.

  7. Runway · Open-sources Praxis-1, a robot control model under Apache 2.0 that leads the World Action Model category on Nvidia's RoboLab-120 but cannot yet plan complex actions alone.

  8. Amazon · Is building a "World Product Graph" indexing products sold anywhere online, including items it doesn't sell, so AI agents rely on its product index.

  9. DeepSeek · Open-sources the Ultrafast inference engine for DeepSeek-V4.1-Flash, supporting 1M-token context and multi-server KV cache sharing, reaching 2,060 tokens/s decode on eight B300 GPUs.

  10. DeepSeek v4.1 flash · Becomes the first open-weight model to score above 1% on the SWE-sweep bug-finding benchmark, still behind Gemini 4 Hydrogen's 13.5% and GPT Astra 6's 9.5%.

View all →

Launches

PH · HN · 13

Agweb

A macOS browser for coding agents built on WebKit, so multiple agents don't get in your way.

▲ 24 · Show HN · 6 comments ·

Gemini Robotics 2.5

Google's most capable vision-language-action robot model, converting sight and instructions into movements across robot types.

▲ 97 · Product Hunt · 2 comments ·

HiringNow

AI job agent that finds and applies to jobs before anyone else, with your approval.

▲ 94 · Product Hunt · 7 comments ·

Dex Semantics

Adds declarations to analytical models so agents build analytics with 25% fewer errors.

▲ 92 · Product Hunt · 9 comments ·

Kokonut UI

Open-source React components for AI coding agents, installable and editable via MCP.

▲ 87 · Product Hunt · 2 comments ·

Tabletop by Google Labs

Build interactive group tools for trips, games, and events with real-time AI collaboration.

▲ 85 · Product Hunt · 2 comments ·

MAP & GO!

Put your tasks and events on a map with weather and one-tap AI navigation.

▲ 83 · Product Hunt · 4 comments ·

Antelog

Watches AI coding agents to prevent data leaks and deletion, and logs every action.

▲ 82 · Product Hunt · 15 comments ·

Businesses in ChatGPT

Lets businesses claim profiles and be discovered by customers inside ChatGPT.

▲ 82 · Product Hunt · 5 comments ·

Valutorio

Native Mac and iPhone Markdown notes app that lets AI read and write via MCP.

▲ 81 · Product Hunt · 9 comments ·

Paxper

Visually create cloud infrastructure and review every change, for developers and DevOps.

▲ 102 · Product Hunt · 21 comments ·

Expo Simulators

Platform-agnostic cloud simulators letting agents install builds, drive them, and verify their work.

▲ 80 · Product Hunt · 3 comments ·

Open Source

GitHub Trending · 14

An open-source AI office SDK bringing editing and collaboration to products and agents.

GitHub Trending ·
★ 24.2k+495 today

An open-source AI coding agent that lives in your terminal for developers.

GitHub Trending ·
★ 30.1k+462 today

A coding-agent skill and proxy that cuts 65% of tokens by talking like a caveman.

GitHub Trending ·
★ 119.8k+227 today

An LLM inference engine for Metal, CUDA and Vulkan that serves models without Python.

GitHub Trending ·
★ 2.1k+188 today

Self-hosted workflows that orchestrate coding agents like Codex with approval gates.

GitHub Trending ·
★ 694+174 today

Agent-User Interaction Protocol that brings agents into frontend applications.

GitHub Trending ·
★ 17.2k+9 today

Keeps coding agents up to date with the latest web best practices, for developers.

GitHub Trending ·
★ 4.8k+1.9k today

OpenAI's Codex Security CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities.

GitHub Trending ·
★ 13.1k+751 today

LLM-powered multi-market stock analysis system with real-time news, decision dashboard, and automated notifications.

GitHub Trending ·
★ 71.6k+256 today

Advanced Windows video player and capture enhancement tool with AI super resolution, DLSS5 enhancement, and frame generation.

GitHub Trending ·
★ 1.6k+256 today

Alibaba's open-source hybrid code review tool combining deterministic pipelines and LLM agents for precise line-level comments.

GitHub Trending ·
★ 50.6k+215 today

Workspace to stay in the loop with agents: plan, annotate, review code, share with team, and send feedback to agents.

GitHub Trending ·
★ 10.4k+108 today

Claude Code skills for equity investors and traders: market analysis, charting, economic calendars, screeners, and strategy development.

GitHub Trending ·
★ 4.2k+69 today

Model Ranking · 24 h

2026-11-14 UTC
01DeepSeek V4.1 FlashDeepSeek · 8.09T tokens29.8%—
02Hy4Tencent · 2.11T tokens7.8%—
03DeepSeek V4.1 FlashDeepSeek · 2.01T tokens7.4%—
04GLM 5.3 FlashZ.ai · 1.33T tokens4.9%—
05GPT-6 LunaOpenAI · 999.3B tokens3.7%▲1
06Step 5 PreviewStepFun · 958.8B tokens3.5%▲1
07Claude Opus 5.5Anthropic · 753.6B tokens2.8%▼2
08MiniMax M3.1 FlashMiniMax · 535.4B tokens2.0%▲3
09MiMo-V2.6-FlashXiaomi · 500.5B tokens1.8%▼1
10ERNIE-5.5-Flash-Preview (free)Baidu · 495.5B tokens1.8%▲2

New Models

2
Ling 3.2 Flash (free)inclusionAI · 1.0M ctx · 11/13$0.00 / $0.00
Gemini 4 HydrogenGoogle · 1.0M ctx · 11/12$1.50 / $7.50batch $0.75 / $3.75

Deals

No funding or M&A detected in the past 24 hours.