Skip to main content
Back

Anthropic releases Haiku 4.5: fast, low-cost coding model

OpenAI adds memory, Sora 2 expands video tools, Google upgrades Veo 3.1, Qwen adds chat memory, Gemma aids cancer resear

October 16, 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 32 sec

Top News

▸ Anthropic expands Claude 4.5 family with lightweight coding model

Encord

▸ Explore the world’s largest open-source multimodal dataset, Encord E-MM1

Signals

▸ OpenAI updates ChatGPT with automatic memory management for Plus and Pro

▸ Google releases Veo 3.1 Fast, delivering improved audio-visual content creation

▸ OpenAI rolls out storyboards and extended video limits in Sora 2

▸ Alibaba’s Qwen adds Chat Memory to personalize interactions with history

▸ Google uses Gemma AI to uncover immune-boosting cancer treatment mechanism

Scout Monitoring

▸ Skip dashboards. Monitor performance in real time with Scout MCP

Trending Repos

▸ Clone-Wars: Discover code for clones like Airbnb, Instagram, Spotify

▸ reflex: Build full-stack web apps entirely in pure Python

▸ claude-code-templates: Prebuilt configurations for Claude Code across languages

Top Lecture

▸ Google launches ADK course on real-time voice agents at DeepLearning.AI

Top News

Anthropic introduces Claude Haiku 4.5, a small-scale model built for efficient coding performance

alpha_signal_image_2

Anthropic introduced Claude Haiku 4.5, its newest small-scale model designed to deliver high coding performance at lower cost and latency. The release expands the Claude 4.5 family with a lightweight option alongside Sonnet and Opus, while keeping feature parity across platforms.

Key Features

  • Nearly frontier coding quality at much lower cost.

  • Outperforms Sonnet 4 in some UI / “computer use” tasks.

  • Designed to support parallel orchestration: Sonnet 4.5 can plan, Haiku 4.5 can execute subtasks.

  • High responsiveness makes it ideal for interactive workflows and rapid prototyping.

  • Priced at $1 / $5 per million input & output tokens.

Model Architecture Overview

  • Built with similar architecture goals as Sonnet models but optimized for speed and cost efficiency.

  • Uses tool-enabled pipelines e.g. bash, file edits, to support practical coding tasks.

  • Can be used in parallel with Sonnet 4.5: Sonnet plans, Haiku executes, for faster end-to-end throughput.

Performance & Benchmarks

  • Hits ~73.3% on SWE-bench Verified with tool use, per Anthropic’s reporting.

  • In agentic coding evaluations, Haiku 4.5 reaches ~90% of the performance of Sonnet 4.5.

  • Runs 4-5× faster than Sonnet 4.5 in many settings while maintaining strong output quality.

  • Safety assessments: shows lower rates of misalignment than Haiku 3.5, and even better alignment metrics compared to both Sonnet 4.5 and Opus 4.1. Anthropic

  • Classified under AI Safety Level 2 (ASL-2), less restrictive than Sonnet or Opus models.

Access & Availability

  • Available today via Claude API, apps, Amazon Bedrock, and Google Vertex AI.

  • Replaces Haiku 3.5 and can act as cost-efficient substitute for Sonnet 4 in many workflows.

  • Developers can immediately call claude-haiku-4-5 via Anthropic’s API.

Try now
alpha_signal_image_3

Encord launches the world’s largest open-source multimodal AI dataset

Looking for more AI training data? The open-source E-MM1 dataset has 107 million groups of images, videos, text, audio and 3D point clouds.

  • Open-Source Dataset: 10x larger than the next closest multimodal dataset

  • Natural Language Search: Find the data you need by text input in Encord

  • 6000+ Hours of Human Annotation: 1 million labels manually added to improve data diversity and accuracy

With this dataset, the Encord ML team also released a custom retrieval model - available open-source for anyone to test.

See the dataset

partner with us

Signals

OpenAI announces ChatGPT feature to manage, search, and reprioritize user memories automatically

3,437 Likes

Google expands Gemini API with Veo 3.1, enhancing video quality, audio, and storytelling tools

7,268 Likes

OpenAI updates Sora 2 with longer video limits and storyboard tools for Pro

1,953 Likes

Alibaba’s Qwen rolls out Chat Memory, storing important details to personalize ongoing conversations and responses

1,085 Likes

Google shows Gemma AI accelerates drug discovery, uncovering existing compounds for immune-system cancer defense

20,435 Likes

Skip Dashboards. Debug in Your Editor With Scout Monitoring’s MCP

Scout Monitoring MCP surfaces errors, slow endpoints, and traces straight in your coding agent. Fix production issues using actual telemetry without context-switching.

Free plan available.

Try Scout MCP today ↗️

Repos

Clone-Wars

31,436 Stars

Clone-Wars indexes over 100 open-source clones and alternatives of popular sites like Airbnb, Instagram, Netflix, and Spotify. It includes source code links, live demos, tech stacks, and GitHub popularity metrics.

reflex

26,139 Stars

This repo maps your Python logic to a React frontend and FastAPI backend under the hood. It uses WebSockets for state sync and lets you deploy apps with minimal setup

claude-code-templates

8,482 Stars

It hosts ready-to-use templates and components to accelerate development with Claude Code. You can set up project scaffolding or individual modules like security auditing agents or GitHub integration.

Top Lecture

Build real-time AI voice agents with Google’s open-source ADK

1,844 Likes

alpha_signal_image_4

This short course shows how to build, debug, and productionize real-time voice agents using Google’s open-source Agent Development Kit (ADK). ADK provides modular components for models, memory, tools, orchestration, and a web interface for tracing agent reasoning.


You will learn to:

  • Build voice agents that take speech input and respond with speech output

  • Manage session state, context, and memory across interactions

  • Add and integrate custom tools and APIs into workflows

  • Implement guardrails with callbacks for safety and reliability

  • Orchestrate specialized agents to research, script, and generate podcasts

  • Use Gemini text-to-speech for multi-speaker audio generation

  • Deploy agents with persistent memory and monitoring on Vertex AI

Start 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