This week the AI industry showed its two faces at once: the product face and the money face. OpenAI shipped a cheaper model built to operate your computer, Apple put faster on-device AI into new Macs, and Moonshot apologized to its own subscribers after a rough model launch. At the same time, Anthropic filed the paperwork for what could be the first IPO of a frontier AI lab, and Nvidia promised to cut AI factory deployment from six months to two weeks. Here is what happened, with every claim verified against primary sources.
Key Takeaways
OpenAI released GPT-6.1 Terra on October 26, a mid-tier model built for computer use, priced at $1 per million input tokens and $5 per million output tokens, 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. (BleepingComputer)
Anthropic filed for a US IPO on October 26, planning to trade on Nasdaq under the ticker ANT, disclosing $16.3 billion in first-half 2026 revenue against a net loss of $80.7 billion. (Reuters)
Moonshot AI apologized for rate limiting and wasted quotas after the Kimi K3.1 Pro launch, offering all subscribers a 25% monthly quota reset. (Reuters)
Apple unveiled M6 MacBook Pro and iMac models on October 27, starting at $1,999 and $1,499, with up to 25% faster AI performance than the M5 MacBook Pro and up to 5x faster on-device AI than the M4 iMac. (Reuters)
Nvidia introduced DSX POD AI factories on October 21, cutting deployment at ready sites from six months to 14 days, with global shipments planned from early 2027. (Nvidia)
1. OpenAI Ships GPT-6.1 Terra: A Cheaper Model Built to Use Your Computer
OpenAI released GPT-6.1 Terra on October 26, positioning it as a fast, efficient mid-tier model that combines detailed visual reasoning and precise computer use with clearer, more natural writing, reports BleepingComputer, citing OpenAI's launch post. The pitch is not raw intelligence but everyday office work: drafting proposals, interpreting technical drawings, filling out forms, and editing video.
The headline is price. Standard API rates are $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens, which OpenAI says is 50% lower than both the previous GPT-5.6 Terra and the larger GPT-6.1 Sol, and roughly one-tenth the token price of the flagship GPT-6 Astra. Terra is available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users, and to developers through the API as `gpt-6.1-terra`, with a 1.05-million-token context window.
On OpenAI's own evaluations, the gains are large. On Agents' Last Exam, a test of professional task execution, Terra scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort while costing about one-third as much per task, and OpenAI says it performs on par with GPT-6 Astra at one-tenth the price per token. On Terminal-Bench 4.0, a coding agent test, Terra solved 50.3% of tasks at maximum effort, up from 21.5% for GPT-5.6 Terra. On its internal error-prompt set, the error rate at low reasoning effort fell from 53.6% to 16.7%. These are the company's own figures, not independent tests, but the direction is consistent: cheaper and more capable at once, with computer use as the headline feature.
The timing matters. Terra arrives the same week Anthropic previewed its proactive Duet agent and Google put OpenAI models in its own cloud store, a story covered in last week's roundup. The race is no longer only about the smartest model. It is about the cheapest model that can reliably do a job on a screen.
2. Anthropic Files for Its IPO: $16.3B in Revenue, an $80.7B Loss, Ticker ANT
Anthropic filed for a US IPO on October 26, planning to trade on Nasdaq under the ticker symbol ANT, reports Reuters. The filing made public the numbers behind one of the most watched private companies in tech, and they are extreme in both directions.
Revenue in the six months to June 30, 2026 was $16.3 billion, up from $1.2 billion a year earlier, a roughly 13.6-fold increase. Adjusted operating income for the period was $1.6 billion. The net loss, however, was $80.7 billion, driven largely by accounting charges tied to convertible notes, as reported the previous week. Compute and infrastructure costs rose to $8.9 billion in the first half, up from $2.3 billion last year, though they fell as a share of revenue to about 55%, from 10 times revenue in 2023.
The filing also laid out the risks in unusual candor. The prospectus warns that advanced AI could exhibit self-preserving behaviors, including attempts to resist shutdown, conceal or manipulate information, and behavior resembling blackmail. Some current and prospective investors are valuing Anthropic at about $1.5 trillion, down from roughly $2 trillion a week earlier. Goldman Sachs and Morgan Stanley were named as lead banks, alongside 20 other underwriters.
Markets reacted fast. AI-linked stocks rose on the day of the filing, with Nvidia adding nearly 3% and Microsoft and Amazon gaining nearly 1% each, reports Reuters. If the listing succeeds, it would be the first stock market debut by a frontier AI lab, and a test of whether public investors will pay a record valuation for a company whose accounting losses are larger than its revenue.
3. Moonshot Apologizes for Kimi K3.1 Pro: A 25% Quota Reset for Every Subscriber
Not every launch goes to plan, and Moonshot AI said so publicly. After the Kimi K3.1 Pro launch, the company apologized for rate limiting and unnecessary quota consumption, reports Reuters. The remedy is unusually direct: Moonshot offered a reset of 25% of every subscriber's monthly quota, described as equivalent to about one week of usage, with subscribers who used less than 25% reset to zero.
Moonshot also said it adjusted parameters to reduce unnecessary subagent calls and output length, the two behaviors blamed for burning through quotas. The episode is a useful counterweight to the week's launch headlines. A model can be technically strong and still fail its users if the serving layer wastes their budget. Moonshot's fix treats quota as a trust issue, not just a billing line, and that framing is worth watching as more labs sell agents that consume tokens on a user's behalf.
4. Apple Unveils M6 Macs: Faster On-Device AI, Same Starting Prices
Apple unveiled a new 14-inch MacBook Pro and iMac on October 27, powered by its latest M6 chip, touting faster performance for running AI tasks directly on the devices, reports Reuters. The 14-inch MacBook Pro starts at $1,999 and the iMac at $1,499. Customers can pre-order from October 28 in 35 countries and regions, including the US, with availability starting November 6.
The numbers Apple cites are specific. The M6 MacBook Pro delivers up to 30% faster graphics and up to 25% faster AI performance than its M5 predecessor, along with up to 24 hours of battery life, the longest ever in a Mac. The new iMac delivers up to 5x faster on-device AI performance than the previous M4 model. The launches come a day before Apple hosts a "Pro and tell" event in Los Angeles on October 28, which Bloomberg News has reported is expected to feature Apple's first touch-screen MacBook Pro.
The strategic read is straightforward. While OpenAI and Anthropic fight over cloud agents, Apple is betting that a large share of everyday AI work will run on the device in front of you, privately and without a round trip to a data center. Faster on-device AI is Apple's answer to the agent era, and the M6 is its latest evidence.
5. Nvidia's DSX POD: AI Factories That Ship in 14 Days Instead of Six Months
Nvidia introduced DSX POD AI factories on October 21 at GTC Berlin, a reference design for a complete Vera Rubin AI factory built by partners inside modular containers, integrated and validated before shipping, reports Nvidia. At sites that are already prepared, deployment time falls from six months to just 14 days. Global shipments by the partner ecosystem are planned from early 2027.
The idea reframes AI infrastructure as a product rather than a construction project. Instead of assembling a data center piece by piece on site, partners ship a pre-validated unit and plug it in. For a week in which Meta planned a $95 billion equity raise and SpaceX disclosed a roughly $40 billion AI chip funding deal, both reported by Reuters, the DSX POD is the supply-side answer: the bottleneck is not only money, but how fast capacity can physically arrive. Fourteen days instead of six months changes what a buildout timeline looks like.
How the Week's Stories Stack Up
Aspect | Story | Key Number | Source |
|---|---|---|---|
Model launch | OpenAI GPT-6.1 Terra | $1/$5 per million tokens, 50% cheaper | BleepingComputer / OpenAI |
Public markets | Anthropic IPO filing | $16.3B revenue, $80.7B loss, ticker ANT | Reuters |
Launch mishap | Moonshot Kimi K3.1 Pro | 25% quota reset for all subscribers | Reuters |
Hardware | Apple M6 Macs | From $1,499, up to 5x on-device AI | Reuters |
Infrastructure | Nvidia DSX POD | 6 months down to 14 days | Nvidia |
The thread across all five: AI is becoming ordinary infrastructure, and ordinary infrastructure gets judged on ordinary metrics. Price per task, quota fairness, battery life, deployment time, and whether a prospectus can survive public scrutiny. Last week was about who AI answers to. This week is about what it costs, in money, in trust, and in time, to run it at scale.
If you like quick video takes on how new open AI projects work, browse the repo signals for 30-second overviews. For last week's governance-heavy roundup, see This Week in AI: October 24, 2026.
Frequently Asked Questions
What is GPT-6.1 Terra?
GPT-6.1 Terra is OpenAI's mid-tier model released on October 26, 2026, built for visual reasoning and computer use: reading screens, filling forms, and automating everyday office tasks. It costs $1 per million input tokens and $5 per million output tokens through the API, 50% less than GPT-5.6 Terra and GPT-6.1 Sol, and is available in ChatGPT Work and Codex for paid plans plus the API as `gpt-6.1-terra`.
What did Anthropic's IPO filing reveal?
Anthropic filed for a US IPO on October 26, 2026, planning to trade on Nasdaq under the ticker ANT. It disclosed $16.3 billion in first-half 2026 revenue, up from $1.2 billion a year earlier, an adjusted operating income of $1.6 billion, and a net loss of $80.7 billion driven largely by accounting charges on convertible notes. Its prospectus also warns that advanced AI could exhibit self-preserving behaviors.
Why did Moonshot reset Kimi subscriber quotas?
Moonshot AI apologized after the Kimi K3.1 Pro launch caused rate limiting and unnecessary quota consumption. It offered every subscriber a reset of 25% of their monthly quota, about one week of usage, and adjusted parameters to reduce unnecessary subagent calls and output length.
What is new in Apple's M6 Macs?
Apple's M6 14-inch MacBook Pro starts at $1,999 and the M6 iMac at $1,499, with pre-orders from October 28 in 35 countries and availability from November 6. Apple cites up to 30% faster graphics and 25% faster AI performance than the M5 MacBook Pro, up to 24 hours of battery life, and up to 5x faster on-device AI than the M4 iMac.
What is Nvidia's DSX POD?
DSX POD is Nvidia's reference design for a complete Vera Rubin AI factory, built by partners in modular containers and validated before shipping. At prepared sites it cuts deployment from six months to 14 days, with global partner shipments planned from early 2027.
Conclusion
This week the AI story moved from capability to cost. OpenAI halved the price of computer-using AI, Anthropic put a $1.5 trillion question mark in front of public investors, Moonshot paid back its users' trust in quota, Apple made on-device AI a selling point rather than a footnote, and Nvidia promised to ship factories like products. The models will keep improving. The companies that win the next phase will be the ones that make AI cheap enough, reliable enough, and fast enough to deploy that nobody has to think about it anymore.
Read this article in Bahasa Indonesia: Berita AI: 28 Oktober 2026.