Google said on 28 October that Gemini 4 Carbon will launch at $2 per million input tokens and $10 per million output tokens, half of Claude Opus 5.5’s listed price, by our arithmetic. Two days earlier, OpenAI said GPT-6.1 Terra matches GPT-6 Astra on one test at one-tenth the price per token.
On 29 October, Mistral AI released the weights of its 1-trillion-parameter Mistral Large 4. Its licence says a licensee that “operates a Model as a Service business” must take a separate commercial licence with Mistral before using the model for any commercial purpose.
Short explainer · Gemini at Half of Opus, Terra at a Tenth of Astra
A vertical news Short summarising this story. Figures come from Google, OpenAI, Anthropic and Mistral AI; our arithmetic is labelled.
Business Pill · Open Weights Explained
An evergreen Business Engineer explainer on what open model weights are. It teaches the general idea only and does not discuss these launches.
The key insight: As we read the week, Google and OpenAI priced closed models below Opus 5.5 and Astra, by our arithmetic, while Mistral opened its weights and attached a licence condition for the businesses that would host them.
Google’s blog post says Carbon ranks first on version 5.0 of the Artificial Analysis Intelligence Index at 59. It says Carbon is going first to a subset of cyber defenders, with access then phased to additional trusted partners through its Fairwind Program.
Google says Carbon will launch at the same introductory price as Argon, $2 per million input tokens and $10 per million output tokens. Anthropic’s pricing page lists Claude Opus 5.5 at $4 and $20.
Listed price per million output tokens on the standard tier for GPT-6 Astra, Claude Opus 5.5, Gemini 4 Carbon (introductory launch price) and GPT-6.1 Terra, and the Mistral Large 4 licence condition for hosting businesses. Ratios are FourWeekMBA arithmetic.
OpenAI Says Terra Matches Astra on One Test at a Tenth
In a post on X, OpenAI said GPT-6.1 Terra “even matches GPT-6 Astra at one-tenth the price per token” on Agents’ Last Exam, which it describes as a test of how well agents use software at work.
OpenAI’s API pricing page lists GPT-6.1 Terra at $1 per million input tokens and $5 per million output tokens on the standard tier. It lists GPT-6 Astra at $10 and $50.
Mistral Opens the Weights and Licenses the Hosting
Mistral’s announcement says “Today, we’re releasing the weights of this improved Mistral Large 4”, and that developers can deploy it on their own infrastructure.
The licence on Hugging Face adds a condition for one kind of user. If a licensee “operates a Model as a Service business”, it “must enter into a separate commercial license with Mistral before using the Model or its derivative works for any commercial purpose”.
The Business Lever: Price per Token, Licence for Hosting
The three moves set the price at different layers. Google priced Carbon below Opus 5.5, and OpenAI priced Terra below Astra, both on closed models. Mistral releases the weights and requires a separate commercial licence from Model as a Service businesses, as the licence defines them.
Put on one ladder, the listed output prices run from $50 per million tokens for GPT-6 Astra to $20 for Opus 5.5, $10 for Carbon and $5 for Terra, by our arithmetic from the three pricing sources.
The Structural Read
As we read Google’s post, its top-ranked model goes first to cyber defenders and will launch at half of Opus 5.5’s listed price, by our arithmetic.
As we read OpenAI’s post and pricing page, a model it says matches Astra on one test is listed at a tenth of Astra’s price.
As we read Mistral’s licence, the weights are open to developers, while Model as a Service businesses, as the licence defines them, must take a separate commercial licence.
Mistral Large 4 licence, Hugging Face
“the Licensee must enter into a separate commercial license with Mistral before using the Model or its derivative works for any commercial purpose”
Three Implications
FOR BUYERSOpenAI lists GPT-6.1 Terra at $1 and $5 per million input and output tokens, against $10 and $50 for GPT-6 Astra.
FOR HOSTS Mistral’s licence requires a separate commercial licence for Model as a Service businesses, and the pages read do not state its price.
FOR ANTHROPICGoogle says Carbon will launch at $2 and $10, and Anthropic lists Opus 5.5 at $4 and $20.
The framework says “Close the SCARCE layer (keep proprietary) + Open the ABUNDANT layer (commoditize)”.
As we read it, Google and OpenAI kept Carbon and Terra closed and priced them below Opus 5.5 and Astra, while Mistral opened its weights and attached a commercial-licence condition for hosting businesses. The framework is the owner’s argument and does not discuss these launches.
Google has not made Carbon generally available, and its $2 and $10 are an introductory price. OpenAI’s parity claim covers one test, Agents’ Last Exam. Mistral’s pages read do not state the commercial licence’s price.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
In one week, Google priced its top-ranked model at half of Opus 5.5, OpenAI said Terra matches Astra on one test at a tenth of the price, and Mistral opened Large 4’s weights while requiring hosting businesses to sign a commercial licence.
A note on sourcing. All facts and figures come from Google’s 28 October blog post, OpenAI’s 26 October post on X and its API pricing page, Anthropic’s pricing page, and Mistral AI’s 29 October announcement and licence, read on 31 October 2026. Price ratios are our arithmetic. We did not contact the companies.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
