Skip to main content
Back

xAI Grok 4.6 Launch , DeepSeek V4-Pro GA , Claude Cowork Sync

xAI ships Grok 4.6 at half price, DeepSeek releases V4-Pro, Pathway solves ARC-AGI for $0.0007, Shadcn open-sources chatbot template

August 13, 2026 6 min read

Hey,

Two frontier models dropped on the same day, both swinging hard on price. Grok 4.6 at $2/$6 per million tokens. DeepSeek V4-Pro at $0.87 output. The race to the bottom is now a race to the bottom while staying competitive.

And the RL paper ties it together nicely: smarter, cheaper training that forgets less. The efficiency gains aren't just in inference anymore.

The real story? Frontier is getting commoditized, fast.
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: 6 min 42 sec

Top News

▸ xAI ships Grok 4.6 at half the price of rival frontier models

Orkes

▸ From Prompt to Production AI Workflows

Top News

▸ DeepSeek ships V4-Pro with smarter agents and 50% off-peak API pricing

Render

▸ Run Cursor's coding agents in your own network

Top News

▸ Anthropic syncs Claude browser sessions across desktop, web, and mobile

Signals

1. New tool strips invisible watermarks from Claude, Gemini, and OpenAI outputs
2. Google Cloud: WPP accelerates humanoid robot AI training 10x using Google Cloud G4 VMs, physical simulation, and NVIDIA tech
3. New paper explains why RL beats SFT for multi-task training by updating 20% of weights vs 93%
4. Pathway's 150M model solves ARC-AGI tasks for $0.0007 each, beating cost-efficiency records
5. Four small design choices can tank long-context performance by up to 47%
6. Shadcn open-sources a minimal chatbot template with one-click Vercel deploy
Top News
xAI ships Grok 4.6 at half the price of rival frontier models
36,532 Likes
xAI ships Grok 4.6 at half the price of rival fron

Grok 4.6 just dropped, and the story here is simple: top-tier performance at half the price of the competition.

The model is built for long, multi-step tasks. Think digging through a codebase, researching a topic across many steps, or taking a rough idea all the way to a working app. It focuses on long-running agents and ambitious interactive and visual work, staying with complex tasks across many steps.

Here is what matters:

  • Pricing: $2/M input, $6/M output tokens. That is roughly 60% below GPT-5.6 Sol's $5/$30.
  • 500K token context window, enough for large codebases.
  • Function calling, structured outputs, web search, code execution, plus text and image input.
  • A faster variant is available at double the price for lower latency.

It is strongest on knowledge work and legal reasoning, with price-to-intelligence as its main pitch. Not the best at everything, but very competitive for the cost.

You can use it now via the API with model string grok-4.6. Cursor and Grok Build are offering 2x included usage for the first week.

Read more
forward →
Presented by Orkes
From Prompt to Production AI Workflows

Modern business processes span APIs, services, and AI models - all demanding reliable workflow coordination. Conductor is an open-source platform that manages these processes, tracking state and handling failures.

This webinar shows how Conductor Skills let AI coding assistants like Claude Code and Cursor:

  • Turn natural language into executable workflows
  • Evolve workflows into agentic systems with AI decisions
  • Add human approvals for safe automation


Join us Wednesday, August 19th.

Register now
partner with us →
Top News
DeepSeek ships V4-Pro with smarter agents and 50% off-peak API pricing
11,029 Likes
DeepSeek ships V4-Pro with smarter agents and 50%

DeepSeek V4-Pro just hit general availability, and the headline is simple: it is now the model you actually call in production, not a preview.

Here is what changed and why it matters:

  • 1 million token context window -- that is an entire codebase or a long multi-turn agent session, no chunking needed.
  • 384K max output -- huge for code generation tasks that produce long files.
  • Flexible reasoning effort: set it to low for quick answers, high for complex agent workflows. You control how much the model thinks before responding.
  • Native Responses API support with one-click Codex setup. The Responses API is OpenAI's format for building coding agents -- DeepSeek now speaks that language natively.
  • Off-peak pricing is 50% cheaper than peak rates, so you can schedule heavy workloads overnight and cut costs significantly.

You can try it now via "Expert Mode" on the web or app, or hit the API directly. Model names are unchanged, so no migration needed.

Read more
forward →
Presented by Render
Run Cursor's coding agents in your own network

Run Cursor's self-hosted coding agents on Render, in your own network, without managing Kubernetes.

One-click deploy, then trigger coding tasks from Slack, Linear, or HTTP. Your code and secrets stay in your infrastructure while Render handles compute and scaling.

Deploy the agent
partner with us →
Top News
Anthropic syncs Claude browser sessions across desktop, web, and mobile
10,292 Likes
Anthropic syncs Claude browser sessions across des

Claude's Chrome side panel used to run in total isolation -- what you did in the browser stayed in the browser, invisible to your other devices. That split is now gone.

The Chrome extension is now a full Cowork session. Here's what travels with you:

  • Conversations save to your history, and skills and connectors work in the browser.
  • Claude can see the page you're on and take actions -- clicking links, typing text, navigating pages, and filling out forms using your existing logins.
  • You can collect invoice details from vendor portals in the browser, then continue the same session on desktop to add local files or compare results.

One real risk: prompt injection, where a malicious website hides instructions inside page content to hijack Claude's actions. Anthropic warns browser agents remain vulnerable -- a separate check reviews consequential actions, and Claude still asks before purchases or sharing personal data.

Max and Team get access today, with Pro rolling out over the coming weeks.

Read more
forward →
Signals
1
New tool strips invisible watermarks from Claude, Gemini, and OpenAI outputs
3,792 Likes
2
WPP accelerates humanoid robot AI training 10x using Google Cloud G4 VMs, physical simulation, and NVIDIA tech
Presented by Google Cloud
3
New paper explains why RL beats SFT for multi-task training by updating 20% of weights vs 93%
487 Likes
4
Pathway's 150M model solves ARC-AGI tasks for $0.0007 each, beating cost-efficiency records
270 Likes
5
Four small design choices can tank long-context performance by up to 47%
183 Likes
6
Shadcn open-sources a minimal chatbot template with one-click Vercel deploy
2,485 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