Skip to main content
Back

Amazon Goes After GPT and Claude

READ MORE

December 3, 2024 4 min read

Hey,

Welcome to AlphaSignal – the most read newsletter by AI developers. 


We bring you the top 1% of news, papers, models, and repos, all summarized to keep you updated on the latest in AI.

IN TODAY'S SIGNAL

Read time: 4 min 27 sec

🎖️ Top News

  • Amazon unveils Nova, a new family of multimodal AI models directly competing with GPT and Claude.


📌 SmallCon


⚡️ Trending Signals


📝 Top Models

  • QwQ-32B-Preview: The smartest open model ever released. Excels in analytical and reasoning tasks.

  • LTX-Video: the first DiT-based video generation model capable of generating high-quality videos in real-time.

  • Microsoft OmniParser: a general screen parsing tool, which interprets/converts UI screenshot to structured format.

🧠 Lecture

If you're enjoying AlphaSignal please forward this email to a colleague. 

It helps us keep this content free.

TOP NEWS

Language Models

Amazon unveils Nova, a new family of multimodal AI models directly competing with GPT and Claude

⇧ 1255 Likes

What's New

Amazon just stepped into the frontier model race with Nova, a family of foundation models that compete directly with GPT-4, Claude, and Llama. The suite includes everything from a fast text model to full video generation capabilities, all accessible through Bedrock API.


These models handle tasks involving text, image, and video inputs, focusing on lowering costs and latency while supporting enterprise-grade workloads. You can use Nova to process large datasets, analyze complex media, and create high-quality generative content.


This matters because Amazon now offers a complete stack from lightweight to frontier models, with pricing that undercuts most competitors. Nova Pro costs roughly 1/3 of GPT-4's price while matching or exceeding its performance on several benchmarks.


Model Lineup & What You Get

Nova splits into 6 models:

  • Nova Micro: Text-only, 128K context, blazing fast at 157 tokens/sec - perfect for high-throughput tasks

  • Nova Lite: Handles text, images, video with 300K context - good for general multimodal work

  • Nova Pro: Advanced reasoning, 300K context, processes 15K lines of code - your heavy lifter

  • Nova Premier: Coming in 2025

  • Nova Canvas: Image generation with editing tools

  • Nova Reel: Video generation with camera controls

The Numbers That Matter

  • Micro: $0.035/1M input, $0.14/1M output tokens
  • Lite: $0.06/1M input, $0.24/1M output tokens
  • Pro: $0.8/1M input, $3.2/1M output tokens

Technical Capabilities
Nova Pro scores 75 on the AI Quality Index, beating GPT-4o's November release.

  • Tops VATEX for video understanding
  • Strong scores on TextVQA for visual QA
  • Leads CRAG for retrieval
  • High TIFA scores for text-image faithfulness
  • Strong showing on Berkeley Function Calling Leaderboard

The API gives you:

  • Async calls for video generation
  • Fine-tuning on text, images, video
  • Function calling for agents
  • RAG through Bedrock
  • Streaming responses
  • Cross-region inference
  • 200+ languages in understanding models

Read more

SmallCon: The Free GenAI Virtual Conference ft. Meta, Mistral, HuggingFace, and more

Join AI leaders from Meta, Mistral, Salesforce, DoorDash, Harvey AI, Nubank, Hugging Face, and more at SmallCon on Dec 11th for deep-dive tech talks, panel discussions, and live demos on the latest tech and trends in GenAI.


You'll learn what it takes to build the GenAI stack of the future and put your SLMs into production.


Sneak peek at the agenda:

  • The Future is Small: Why Apple is Betting Big on Small Models

  • Unlocking Enterprise Transformation with GenAI

  • Trends in GenAI: What's New in SLM Training and Serving

  • AI Agents that Work: Lessons Learned from Salesforce

  • Applied AI and Real World Use Cases

  • Data 2.0: Synthetic Data Generation Best Practices

Get your spot

partner with us

TRENDING SIGNALS

AI Agents

ElevenLabs adds AI agents that can speak in minutes with low latency, full configurability, and seamless scalability

⇧ 1547 Likes

Decentralization

Nous Research announces the parallel pre-training of a 15B parameter language model over the internet (watch it train live)

⇧ 2110 Likes

AI Safety

Anthropic creates a Fellows program to help engineers and researchers transition into doing frontier AI safety research

⇧ 1601 Likes

Enterprise

Cohere releases a new model that can precisely search complex enterprise data like long documents, emails, tables, and code

⇧ 882 Likes

LLM in Production

ZenML unveils the LLMOps Database: a no-BS database of how companies actually deploy LLMs in production

⇧ 1561 Likes

TOP MODELS

Open LLM

QwQ-32B-Preview

⇧ 33,191 Downloads

The smartest open model ever released, it solves complex problems in math and programming, excels in analytical and reasoning tasks. It achieves impressive results on benchmarks like GPQA (65.2%), AIME (50%), MATH-500 (90.6%), and LiveCodeBench (50%).

Agents

LTX-Video

⇧ 3210 Downloads

LTX-Video is the first DiT-based video generation model capable of generating high-quality videos in real-time. It produces 24 FPS videos at a 768x512 resolution faster than they can be watched. Trained on a large-scale dataset of diverse videos, the model generates high-resolution videos with realistic and varied content. We provide a model for both text-to-video as well as image+text-to-video usecases

Parsing

Microsoft OmniParser

⇧ 10,455  Downloads

OmniParser is a general screen parsing tool, which interprets/converts UI screenshot to structured format, to improve existing LLM based UI agent. Training Datasets include: 1) an interactable icon detection dataset, which was curated from popular web pages and automatically annotated to highlight clickable and actionable regions, and 2) an icon description dataset, designed to associate each UI element with its corresponding function.

TOP LECTURE

GenAI

Google's 5-Day Gen AI Intensive Course

⇧ 2733 Likes

A comprehensive guide to mastering Generative AI concepts and tools. Originally a live event, it's now available as a self-paced resource to help you build practical AI expertise.


What you'll learn:

  • Day 1: Foundational models, LLM evolution, prompt engineering, and inference optimization.

  • Day 2: Embeddings, vector databases, search algorithms, and real-world applications.

  • Day 3: Building generative AI agents, function calls, and LangGraph integration.

  • Day 4: Creating domain-specific LLMs, grounding with external data, and Gemini model fine-tuning.

  • Day 5: Adapting MLOps for Generative AI, deploying with Vertex AI tools.

Includes assignments, whitepapers, code labs, and livestream recordings for practical learning.

Read more

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 200,000+ AI developers? Let's work together.

Work with us