- AI Adopters
- Posts
- Amazon’s AI reasoning model
Welcome back.
Amazon is about to make a big move in AI with Nova, its upcoming reasoning model. With hybrid thinking and a focus on efficiency, they’re aiming to challenge the biggest names in AI.
But will it stand out in an increasingly crowded field? Let’s take a look.
In today’s release:
1. Amazon to launch an AI reasoning model
2. Opera announces ‘Browser Operator’
3. Deepgram launches Nova-3
Amazon to launch Nova AI model with advanced reasoning
Amazon is set to launch Nova AI reasoning model in June 2025, aiming to compete with ChatGPT, Meta AI, and Google’s Gemini.
The model will take a hybrid reasoning approach, balancing quick responses with deeper, extended thinking for complex problem-solving.
Competitive positioning: Amazon wants Nova to rank in the top 5 on coding and math benchmarks like SWE and AIME.
Anthropic’s approach: Nova's hybrid reasoning follows a similar approach to Claude 3.7 Sonnet, developed by Amazon-backed Anthropic.
Cost focus: Amazon is prioritizing competitive pricing to challenge premium AI services.
The AGI division, led by Rohit Prasad, is tasked with building the model, signalling Amazon’s broader AI ambitions. CEO Andy Jassy has hinted that decreasing AI inference costs could make advanced AI more accessible and will benefit both customers and Amazon's business.
How can you use this to your advantage?
With Nova’s hybrid reasoning, businesses may receive a cost-effective AI solution for automation, coding, and problem-solving. If Amazon delivers on its promise, Nova could offer a strong alternative to existing premium AI models, reshaping how AI is used in customer service, logistics, and decision-making.
Opera unveils ‘Browser Operator’ for AI agentic browsing
Opera has announced Browser Operator, an AI agent designed to autonomously perform browsing tasks for users.
This marks a major shift from traditional web browsing, turning the Opera browser into an active assistant rather than just a display engine.
Key features: Autonomy, decision-making, perception, action execution, and learning.
Task automation: Users can request actions in natural language, e.g., “Buy pink Nike running shoes, size 8.5”, and the Browser Operator will handle it.
Privacy and efficiency: Runs natively inside the browser, accessing pages directly without relying on external cloud processing.
Unlike AI copilots that simply assist with searches, Browser Operator can execute real-world tasks while keeping users in control. It doesn’t rely on pixel-based navigation but instead processes web pages using DOM Tree and browser layout data, making it faster and more efficient.
How can you use this to your advantage?
Once available, Browser Operator could save you time by automating tasks like shopping, bookings, or research. This AI browsing approach could streamline workflows, making complex online tasks more seamless and hands-off.
Deepgram launches Nova-3 medical for AI healthcare transcription
Deepgram has introduced Nova-3 Medical, a specialized AI speech-to-text (STT) model designed to reduce transcription errors in healthcare settings.
Built to integrate with electronic health records (EHRs), Nova-3 Medical is engineered to accurately capture complex medical terms and minimize transcription errors that can impact patient care.
Industry-leading accuracy: 3.45% Word Error Rate (WER), a 63.6% reduction in errors vs. competitors.
Medical keyword precision: 6.79% Keyword Error Rate (KER), reducing errors in critical terms like drug names and conditions by 40.35%.
Enterprise-ready: HIPAA-compliant, deployable on-premises or via Virtual Private Cloud (VPC) for data security.
Nova-3 Medical is 5-40x faster than many competing STT models, making it ideal for real-time transcription in digital health applications. With pricing starting at $0.0077 per minute, it is also a cost-effective alternative to major cloud providers, helping healthcare tech firms scale operations while keeping costs low.
How can you use this to your advantage?
If you're in healthcare, Nova-3 Medical can streamline documentation, reducing manual transcription work and improving EHR integration. For health tech developers, its customization options (up to 100 key medical terms) enable tailored AI solutions for specialized medical fields, enhancing accuracy and workflow efficiency.
OTHER AI NEWS
OpenAI’s NextGenAI: OpenAI launches NextGenAI, a $50M initiative uniting 15 institutions to advance AI in research and education. The program provides funding, compute, and API access to support students, educators, and researchers.
Apple unveils M3 iPad Air: The new iPad Air, powered by M3, boasts up to 2x faster performance than the M1 version. A redesigned Magic Keyboard and Apple Intelligence support round out the upgrade.
Auxia raises $23.5M for AI marketing: This move aims to help enterprises escape the “reacquisition treadmill,” where brands repeatedly pay to win back the same customers. Rising ad costs and reliance on Google, Meta, and Amazon are pushing businesses to prioritize AI for customer retention.
Salesforce launches AI agent marketplace: Salesforce’s AgentExchange debuts with 200+ partners, aiming to automate workplace tasks. Targeting a $6T digital labour market, it offers no-code AI agents for businesses.
Scrunch AI raises $4M for AI search: Scrunch AI helps companies optimize for AI-driven search as chatbots reshape web traffic. With this funding, the company aims to give businesses control over how AI presents them.
POPULAR AI TOOLS
Pieces Long-Term Memory Agent: The first AI that remembers everything you work on.
Greta: Ship any full-stack applications within seconds.
One Task X: One task today > ten half-done.
Snapdeck beta: Generate presentation slides in seconds with AI.
SSSmodel: Prompt 2 models at the same time.
AND THAT’S A WRAP
Thank you for reading!
If you found this email useful, share it with a friend or colleague who also loves AI.
Also, drop me a follow on Twitter/X for more AI and tech updates.
I will talk to you soon!
Mike