AI & Technology

November 2025: The Month That Redefined AI Foundation Models

From Google's record-breaking Gemini 3 to Anthropic's Claude Opus 4.5 and OpenAI's GPT-5.1, November 2025 saw an unprecedented wave of AI foundation model releases that are reshaping the industry.

Curated by Matt Perry

CTO

6 December 2025

A Landmark Month for AI

November 2025 will be remembered as a watershed moment in artificial intelligence. Within just three weeks, the world's leading AI companies released major updates to their foundation models, each pushing the boundaries of what AI can achieve. Here's a comprehensive look at the models that defined this remarkable month.

Google Gemini 3: Setting New Standards

Released: 18 November 2025

Google kicked off the month's major releases with Gemini 3, which the company describes as "a new era of AI intelligence." The model achieved an unprecedented score of 1501 on the LMArena leaderboard, surpassing its predecessor Gemini 2.5 Pro's score of 1451.

Key benchmarks:

91.9% on GPQA Diamond (PhD-level science questions)
37.5% on Humanity's Last Exam
23.4% on MathArena Apex (new state-of-the-art)
81% on MMMU-Pro and 87.6% on Video-MMMU

Gemini 3 introduced several groundbreaking features:

Deep Think Mode: An enhanced reasoning mode that achieves 45.1% accuracy on ARC-AGI-2, demonstrating significant advances in genuine reasoning ability.

Generative UI: Perhaps the most innovative feature, allowing Gemini to create custom user interfaces on the fly for specific tasks like mortgage calculators or physics simulations.

Google Antigravity: A revolutionary agentic development platform that enables AI agents to work autonomously across editors, terminals, and browsers.

API pricing starts at $2 per million input tokens and $12 per million output tokens.

OpenAI GPT-5.1: Two Modes, Maximum Flexibility

Released: 12 November 2025

OpenAI's GPT-5.1 release introduced a dual-mode approach with "Instant" and "Thinking" variants, giving users flexibility based on their needs.

GPT-5.1 Instant is designed for everyday interactions, offering:

Warmer, more natural conversational tone
Improved instruction-following capabilities
Better performance for straightforward queries

GPT-5.1 Thinking excels at complex problem-solving with:

Extended reasoning capabilities for difficult tasks
More persistent problem-solving on complex challenges
Clearer explanations of its reasoning process

OpenAI also introduced new tone presets: Professional, Candid, and Quirky, allowing users to customise how ChatGPT communicates.

Anthropic Claude Opus 4.5: The Coding Champion

Released: 24 November 2025

Anthropic's Claude Opus 4.5 arrived as the month's final major release, positioning itself as the premier choice for developers and enterprises requiring advanced coding and agentic capabilities.

Standout achievements:

State-of-the-art performance on SWE-bench Verified (software engineering benchmark)
Best-in-class results for agentic tasks and computer use
New "effort" parameter in the API for adjustable compute allocation

Anthropic emphasised safety improvements alongside capability gains, with enhanced guardrails and monitoring for autonomous operations.

Pricing is set at $5 per million input tokens and $25 per million output tokens, reflecting its position as a premium offering for complex tasks.

xAI Grok 4.1: The Silent Contender

Released: 17 November 2025

xAI took a different approach with Grok 4.1, rolling out the model quietly between 1-14 November before the official announcement on November 17. Available in "Auto" mode, Grok 4.1 demonstrated competitive performance across benchmarks whilst maintaining xAI's focus on real-time information access through X (formerly Twitter) integration.

What This Means for Businesses

The rapid succession of releases creates both opportunities and challenges for businesses looking to leverage AI:

Increased competition: With multiple capable models available, businesses have more choice but also face more complexity in evaluation.

Specialisation emerging: Each model now has distinct strengths, from Gemini's multimodal capabilities to Claude's coding prowess to GPT's conversational flexibility.

Pricing pressure: Competition is driving more competitive pricing, making advanced AI more accessible to smaller organisations.

Integration considerations: Businesses may benefit from multi-model strategies, using different AI systems for different tasks.

Looking Ahead

November 2025's releases signal that we're entering a new phase of AI development where the focus is shifting from raw capability improvements to practical applicability. Features like generative UI, adjustable effort parameters, and tone presets show that AI companies are increasingly focused on making their models more useful in real-world contexts.

For businesses and developers, the message is clear: the AI landscape is evolving rapidly, and staying informed about these developments is essential for making strategic technology decisions.

Sources

Photo credit: geralt on Pixabay

This article was compiled from the following sources:

Ready to put AI to work in your business?

Book a free 30-minute discovery call. We will discuss your goals, identify quick wins, and outline a practical plan to get started.

Book a discovery call

Frequently Asked Questions

Which AI model from November 2025 is best for my business?

It depends on what you need it to do. Google Gemini 3 is strongest for multimodal tasks involving images, video, and text together. Claude Opus 4.5 leads the field for software engineering and complex coding work. GPT-5.1 offers the most flexible conversational experience with its Instant and Thinking modes. Many businesses find the best approach is using different models for different tasks rather than committing to just one.

How much do these AI foundation models cost to use?

API pricing varies across providers. Google Gemini 3 starts at $2 per million input tokens and $12 per million output tokens, making it one of the more affordable options. Claude Opus 4.5 sits at the premium end with $5 input and $25 output per million tokens. For businesses using ChatGPT or Claude through their consumer products rather than the API, subscription plans typically range from £16 to £20 per month per user. The increased competition between providers is steadily pushing prices down.

Do I need to switch AI providers every time a new model comes out?

No. Chasing the latest model every few weeks is neither practical nor necessary for most businesses. What matters more is choosing a provider that fits your workflow and building your processes around it. That said, it is worth reviewing your AI tools every 6 to 12 months to make sure you are still getting the best value. The November 2025 releases showed that meaningful improvements are happening quickly, so a model that was the clear leader six months ago may no longer be the best fit.

What does the rise of "agentic" AI features mean in practical terms?

Agentic AI refers to models that can take actions on your behalf, not just answer questions. For example, Google's Antigravity platform lets AI agents work across code editors, web browsers, and terminals autonomously. Claude Opus 4.5 can operate computer interfaces to complete multi-step tasks. In practical terms, this means AI is moving from being a tool you ask questions to, towards being an assistant that can carry out entire workflows with minimal supervision. For businesses, this could mean automating research, data entry, testing, and reporting tasks that currently require human attention.

Get AI insights like this every week

Subscribe Free