Skip to main content
Back

DeepSeek DeepInfer , Claude Duet Agent , EAGLE-4 961 tok/s

DeepSeek open-sources inference engine, Anthropic previews Claude Duet, Unitree demos kung fu humanoid, OpenAI finds PHP flaw

October 22, 2026 7 min read

Hey,

Inference speed is having a moment. DeepSeek just open-sourced its actual production engine. FlashMap hits 6x faster inference. EAGLE-4 clocks 961 tokens/s. The whole stack is getting faster, cheaper, and more open.

Meanwhile, Anthropic ships Claude Duet, a proactive agent that works while you sleep. Faster inference enables exactly this: always-on agents that don't burn your budget.

Speed unlocks behavior. That's the real story today.
In Partnership with · partner_image
Author photo
Today's Author
Lior Alexander. Founder of AlphaSignal and former ML Engineer. Previously built ML systems at Iguazio, Guesty, Enphase, Mila.
Summary

Read time: 7 min 48 sec

Top News

▸ DeepSeek open-sources its production inference engine to cut compute costs

Fireworks.ai

▸ Join Forge by Fireworks: Nov 3, San Francisco

Top News

▸ Anthropic previews Claude Duet, a proactive agent that works while you're away

Neo4j

▸ Build AI with Better Context

Top News

▸ Unitree unveils a kung fu humanoid robot with 8-hour battery life

Signals

1. OpenAI models find critical security flaw in PHP, which powers 70% of websites
2. Pave: Connect custom apps to 350+ tools in your stack, turning the apps you already use into ones that work for you
3. Nebius drops open cookbook for building AI apps with open-source models
4. New speculative decoding method hits 6x faster inference on Qwen3-8B
5. DeepSeek V4.1 Flash gets a 339 GB GGUF quantization that runs on consumer hardware
6. EAGLE-4 hits 961 tokens/s, up to 4x faster than vanilla LLM inference
Our Events
Requirements, Challenges, and Solutions for Full-Duplex Model Training
Oct 27, 10a PT · Online
Voice AI that can listen and talk at the same time is the next frontier, and training those full-duplex models is a whole different problem. Join Sully Goolamnobee, Product Manager for AI Voice at Voices.com, on Tue, Oct 27 at 10:00 AM PT for a live session on the requirements, challenges, and solutions for full-duplex model training, with live Q&A at the end.
Designing Flexible, Efficient AI Infrastructure with Google Cloud's AI Hypercomputer
October 29, 10a PT · Online
Scaling AI shouldn't mean choosing between raw performance and cost efficiency. Join Chelsie Czop, Sr. Product Manager for AI Infrastructure at Google Cloud, on Thu, Oct 29 at 10:00 AM PT to see how AI Hypercomputer, NVIDIA GPUs, and smart scheduling keep training and inference fast, resilient, and affordable, with live Q&A at the end.
Building a Semantic Backbone for Agentic AI over Enterprise Data
Nov 5, 10a PT · Online
AI agents are only as smart as the enterprise data they can actually understand, and most of that data is scattered and siloed. Join Vassil Momtchev, SVP of Product & Technology at Graphwise, on Thu, Nov 5 at 10:00 AM PT to learn how to build a semantic backbone for agentic AI, with a live demo of the Graphwise AI Suite and Q&A at the end.
Top News
DeepSeek open-sources its production inference engine to cut compute costs
49,039 Likes
DeepSeek open-sources its production inference eng

DeepSeek just open-sourced DeepInfer, the actual engine running their models in production. Not a demo. Not a research paper. The real thing.

Think of an inference engine as the software layer that takes a trained AI model and makes it answer questions fast, at scale, without melting your server budget. That is what DeepInfer does.

Here is what is inside:

  • Speculative decoding via DSpark: a draft model guesses tokens ahead, a confidence filter trims ~30% of checks, netting a 4-7% throughput gain
  • Distributed inference to split work across multiple chips
  • Structured outputs so responses follow a set format
  • Runs on Nvidia SM100 and Huawei Ascend 950 chips

They also dropped deepseek-recipe, a token encoding library that mirrors the official DeepSeek API exactly. Self-host it and your outputs behave identically to the hosted version. No more subtle mismatches.

Both repos are on GitHub under Apache 2.0. Future model code lands here too.

Read more
forward →
Presented by Fireworks.ai
Join Forge by Fireworks: Nov 3, San Francisco

The most ambitious companies are training specialized models to outperform the frontier on their domain.

Join Fireworks CEO Lin Qiao, NVIDIA CEO Jensen Huang, and leaders from Microsoft, Replit and more at Forge on November 3 in San Francisco.

One day dedicated to building your own frontier, with keynotes, technical sessions, and conversations with the builders pushing specialized intelligence forward.

Apply to attend in person or join the livestream virtually.

Explore forge
partner with us →
Top News
Anthropic previews Claude Duet, a proactive agent that works while you're away
42,516 Likes
Anthropic previews Claude Duet, a proactive agent

Most AI tools sit there waiting for you to type something. Claude Duet flips that. It comes to you.

Anthropic just launched a research preview of Duet, a standalone app where Claude acts as a proactive agent. Think of it less like a chatbot and more like a background worker that monitors your apps, surfaces what matters, and handles tasks while you are away.

Here is what it actually does:

  • Sends you a morning brief from your connected apps
  • Monitors email, calendars, and Slack, then pings you only when something needs you
  • Connects to 6,000+ tools including invoicing and scheduling
  • Keeps running in the cloud even when your device is off
  • Asks your approval before sending messages, payments, or deleting anything

It runs on Claude Opus 5.5 and pulls in your existing Claude memory and connected apps, so setup is minimal. It never sees your passwords, and a dedicated safety layer guards against leaking sensitive info.

Access is via waitlist for Pro and Max subscribers in the US, at no extra cost, though background tasks count toward your usage limits.

Read more
forward →
Presented by Neo4j
Build AI with Better Context

Scattered data, missing context, and limited memory undermine AI agents’ reasoning. NODES 2026 gathers engineers using context graphs and agentic memory to fix that, in a 24-hour online event on November 12

  • Fireside Chat with Andrew Ng: What’s Next for AI Builders?
  • 100+ live technical sessions
  • Online, 24 hours, all time zones
Save your spot
partner with us →
Top News
Unitree unveils a kung fu humanoid robot with 8-hour battery life
20,309 Likes
Unitree unveils a kung fu humanoid robot with 8-ho

Unitree just dropped the H2-X Kung Fu Legend Guardian, a humanoid robot built for security and rescue. Think less warehouse helper, more bodyguard with a black belt.

Here is what it can actually do:

  • Run for 8 hours on a single charge, nearly triple the base H2's 3-hour battery life
  • Jump off 4-meter platforms and land without falling over
  • Lift cars to free people trapped underneath
  • Execute real martial arts moves using an AI model that predicts opponent actions in real-time, no human controlling it
  • Pack 2,070 TOPS of onboard AI compute, roughly the processing muscle needed to run complex decisions locally

It stands 182cm tall, weighs 70kg, and has 31 movable joints. Unitree actually warns you to stay 2-3 meters away while it operates. That tells you everything about how much force those joints generate.

No pricing or release date announced yet. The base H2 starts at $29,900 for context.

Read more
forward →
Signals
1
OpenAI models find critical security flaw in PHP, which powers 70% of websites
13,781 Likes
2
Connect custom apps to 350+ tools in your stack, turning the apps you already use into ones that work for you
Presented by Pave
3
Nebius drops open cookbook for building AI apps with open-source models
269 Stars
4
New speculative decoding method hits 6x faster inference on Qwen3-8B
521 Likes
5
DeepSeek V4.1 Flash gets a 339 GB GGUF quantization that runs on consumer hardware
23,033 Downloads
6
EAGLE-4 hits 961 tokens/s, up to 4x faster than vanilla LLM inference
8,470 Likes
At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 300,000 developers stay informed and ahead. We're passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.

Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We're based in the United States. By using our site, you agree to be governed by U.S. laws.
Looking to promote your company, product, service, or event to 300,000+ AI developers?
Work with us