Skip to main content
Back

xAI releases Grok 4 Fast: beats Claude 4.1, Gemini 2.5

Google brings Gemini to Chrome, Alibaba ships Qwen-Next FP8, Stanford debuts Paper2Agent, OpenAI and Gemini shine at ICP

September 22, 2025 4 min read

Hey

Welcome to AlphaSignal, the most read source of news by AI engineers and researchers.


Every day, we identify and summarize the top 1% of news, papers, models, and repos, so you’re always up to date.


Here’s today’s roundup:

Summary

Read time: 4 min 23 sec

Top News

▸ xAI releases Grok 4 Fast, a cheaper multimodal reasoning model

Signals

▸ Google brings Gemini AI to Chrome for smarter, safer browsing

▸ Alibaba unveils FP8 builds of its new open reasoning models for faster inference

▸ Stanford announces an open-source tool that makes research papers interactive

▸ OpenAI reports first place at ICPC with 12/12 problems, top coding competition

▸ Google’s Gemini earned ICPC gold by solving 10/12 programming problems

Trending Tutorials

▸ OpenAI shows how to use Codex CLI locally in terminal for coding

▸ LangChain's tutorial on turning Claude Code into domain specific agent

▸ Build a YouTube research agent in 15 minutes using Claude Code

AI Coding Tip

▸ Automate repetitive tasks by letting Copilot agents handle issues

Top News

xAI launches Grok 4 Fast: frontier accuracy using 40% fewer thinking tokens, lowering compute costs

alpha_signal_image_2

xAI has launched Grok 4 Fast, a multimodal reasoning model that delivers near-frontier performance while being 98% cheaper to run than Grok 4.


Grok 4 Fast shows that frontier-level intelligence can run at a fraction of the cost, making large-scale reasoning models more accessible and affordable for everyday use.


Key Features

  • Uses 40% fewer thinking tokens while keeping accuracy high

  • Cuts compute cost by 98% compared to Grok 4

  • Supports multimodal inputs (text + images)

  • Built-in tool use for browsing and code execution

  • Available in Fast and Auto modes across platforms

  • Handles 2M token context for long reasoning tasks

How it works

It uses fewer “thinking tokens” while keeping reasoning quality, cutting compute costs.

  • Optimizes inference steps for faster responses at lower cost

  • Maintains accuracy by refining reasoning chains instead of brute-force token expansion

  • Uses xAI’s tool integrations browsing, code execution for multi-step reasoning

Performance

Normally, high-performing models require heavy compute, but Grok 4 Fast achieves comparable results using 40% fewer thinking tokens.

  • Hits 85.7% on GPQA Diamond (science) and 92% on AIME 2025 (math)

  • Ranked #1 in LMArena’s Search Arena

  • On coding tasks it even outperformed the larger Grok 4

  • Beats Claude 4.1 Opus and Gemini 2.5 Pro across several categories

Access and Availability

  • Free on grok.com, grok.x.com, iOS, and Android

  • Free for a limited time on OpenRouter and Vercel AI Gateway

  • API pricing: $0.20 / 1M input tokens, $0.50 / 1M output tokens

Try now

Signals

Google integrates Gemini directly into Chrome with multi-tab context, AI Mode, and safer browsing

9,357 Likes

Alibaba releases FP8 builds of Qwen-Next, an efficient 80B open-source long-context reasoning system

1,043 Likes

Stanford introduces Paper2Agent, an open tool to transform static research papers into interactive conversational assistants

717 Likes

OpenAI announces its reasoning system outperformed all human teams at ICPC 2025 finals by solving 12/12 problems

5,538 Likes

Google’s Gemini 2.5 Deep Think solved 10 of 12 problems at ICPC World Finals, earning gold-medal level

7,249 Likes

Tutorials

OpenAI shows how to use Codex CLI locally in terminal for coding

2,585 Likes

This guide shows how Codex CLI runs locally as an open-source coding agent. It lets you install via npm or brew, authenticate via ChatGPT, and interactively read, modify, and run code in your project directory with model selection.

LangChain's tutorial on turning Claude Code into domain specific agent

535 Likes

The article explores how to optimize Claude Code for niche libraries. Key lessons include designing clear domain instructions, using MCP for remote documentation access, and balancing quality vs cost when customizing coding agents.

Build a YouTube research agent in 15 minutes using Claude Code

871 Likes

You learn step-by-step how to design and deploy a Claude Code research agent for YouTube. The guide includes permissions, spec writing, slash command creation, and batch workflows for analyzing channels and extracting creator strategies.

AI Coding Tip

Automate repetitive tasks by letting Copilot agents handle issues

You can offload repetitive tasks to GitHub Copilot agents by assigning them directly to issues.


How to use it:
In the chat thread, type:


Assign yourself to this issue and draft a fix

What happens:

  • The agent analyzes the codebase

  • Identifies the root cause

  • Drafts a pull request with the fix

  • Updates progress in the task pane

When to use
For routine bugs, documentation updates, and dependency upgrades.

Benefits
Frees you from repetitive work and keeps you focused on complex tasks.

Read now

At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 200,000 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 250,000+ AI developers?

Work with us