AI
AI Got a Price Tag for Real Work This Week, and the Intro Rates Expire November 30
The frontier labs stopped selling answers this week and started selling shifts. Google opened Gemini 4 Argon to developers and businesses worldwide on October 13, priced at $2 per million input tokens and $10 per million output, with the intro rates doubling on December 1. Meta's Muse Wonder lists at $4 per million input tokens. Mistral has promised open weights for its next large model by the end of October. Apple pushed AI into the living room with HomeView. The race is now about who prices long running agent work the best, and every lab is showing its math.
Argon is Google's answer to the agent era, built for deep work across coding, finance, law, cybersecurity and science, the multi step tasks that run for hours rather than seconds. Reuters reported the worldwide release after months of delays and a leadership shakeup that saw DeepMind cofounder Demis Hassabis step aside. Teams testing Argon in Google's Antigravity coding tool increased agentic code submissions by more than 35% while cutting code rollbacks by nearly a third, with early customers including Deloitte, LexisNexis, Ambience Healthcare and Isomorphic Labs. The pricing undercuts Anthropic's Claude Opus 5.5 at $4 and $20 and OpenAI's GPT-6 Astra at $10 and $50, though the gap closes on December 1 when Argon's rates double.
The pricing page is now the product page. Meta released Muse Wonder at $4 per million input tokens in early October while keeping its full Wonder tier for Muse Code subscribers. Mistral's answer is the opposite of a price cut, downloadable weights for its next large model by October's end, which would put frontier capability in the hands of university labs and small manufacturers for the cost of their own hardware. Google's own tests showed Argon generated practice data lifting a real robot's success rate from 56% to 75%. The pattern is set, top tier models reach paying subscribers and vetted lists first, and each new seller pushes the price of intelligence down for everyone behind them.
Google also gave its assistant a job description. A new task mode in the Gemini app lets the assistant keep working while the user steps away from the computer, connected to more than 40 applications including Instacart, OpenTable, Adobe and Shopify. The pitch is a bounded workflow with approval before external actions, a code migration or research task that runs in the background and checks in when it needs a human. Builders should measure total task cost, tool calls, retries and human review, because the token price is only the first line of the invoice. The agent era will be won on reliability per dollar, and this is the first week the receipts are public.
Apple took the same logic into the home. At its Welcome Home event on October 13, the company introduced HomeView, a $299 smart home hub with a 7 inch Liquid Retina display, a new Siri AI, and facial recognition for up to six household members. Preorders opened immediately, with availability starting November 13, and the wall mounted version needs mounts sold separately. The device manages doorbells, thermostats and cameras alongside music and FaceTime, and personal context features pair with an Apple Intelligence device. Where the labs sell shifts to developers, Apple is selling a presence to families, and both bets assume the same thing, that AI earns its keep by doing work.
For readers, the practical read is simple. Builders get the cheapest frontier window in months, with Argon's intro pricing live through November 30 and open weights arriving from Mistral. Buyers should watch output per dollar, because the labs have started publishing the numbers that matter. Households get a shared screen that knows six faces and keeps each person's context separate. The shift based economy of intelligence has its price list now. Spend it on tasks with receipts, and let the chatbots keep the small talk.
Quick answers
What is this story about?
The frontier labs stopped selling answers this week and started selling shifts. Google opened Gemini 4 Argon to developers and businesses worldwide on October 13, priced at $2 per million input tokens and $10 per million output, with the intro rates doubling on December 1. Meta's Muse Wonder lists at $4 per million input tokens. Mistral has promised open weights for its next large model by the end of October. Apple pushed AI into the living room with HomeView. The race is now about who prices long running agent work the best, and every lab is showing its math.
Why does this story matter?
For readers, the practical read is simple. Builders get the cheapest frontier window in months, with Argon's intro pricing live through November 30 and open weights arriving from Mistral. Buyers should watch output per dollar, because the labs have started publishing the numbers that matter. Households get a shared screen that knows six faces and keeps each person's context separate. The shift based economy of intelligence has its price list now. Spend it on tasks with receipts, and let the chatbots keep the small talk.
Sources
New to crypto? Read the crypto glossary, browse frequent questions, read our story, or explore the story archive.