← 목록으로
AI

OpenAI’s ~$1 Billion GPT-6 Astra Heralds The Age Of AGI, According To NVIDIA’s Jensen Huang, Yet Anthropic’s Fable 5.1 Still Beats It On SWE-Bench Pro And Coding Benchmarks

· 읽기 16분 · 2회

🔗 원문 전체 보기 → Wccftech

OpenAI’s ~$1 Billion GPT-6 Astra Heralds The Age Of AGI, According To NVIDIA’s Jensen Huang, Yet Anthropic’s Fable 5.1 Still Beats It On SWE-Bench Pro And Coding Benchmarks

요약

OpenAI is in full-on hype mode following the release of the GPT-6 Astra model, with NVIDIA's Jensen Huang also having jumped onto the hype-heavy bandwagon just a few hours ago. And yet, as outlandish as Astra's 100,000-GPU training run sounds, it is all set t…

본문

AI Hardware OpenAI’s ~$1 Billion GPT-6 Astra Heralds The Age Of AGI, According To NVIDIA’s Jensen Huang, Yet Anthropic’s Fable 5.1 Still Beats It On SWE-Bench Pro And Coding Benchmarks Rohail Saleem • Sep 7, 2026 at 10:53am EDT Add Wccftech on Google Comments
A person in a black jacket holds a computer chip with a visible X-shaped pattern at the center.
NVIDIA's Jensen Huang has also jumped onto the hype-heavy bandwagon

OpenAI is in full-on hype mode following the release of the GPT-6 Astra model, with NVIDIA's Jensen Huang also having jumped onto the hype-heavy bandwagon just a few hours ago. And yet, as outlandish as Astra's 100,000-GPU training run sounds, it is all set to be eclipsed by an even more ambitious 400,000-GPU training run that might span the entirety of OpenAI's Stargate project in Abilene, Texas.

OpenAI might have spent between $500 million and $1 billion to train the GPT-6 Astra, but an upcoming training run is all set to break Astra's record if Jensen Huang's projection is anything to go by

Last week, OpenAI revealed that it had trained the GPT-6 Astra on a 100,000-GPU cluster. This gigantic training run is likely to have cost the high-flying AI lab anywhere between $500 million and $1 billion, assuming $2.50 - $3.50 in per GPU hourly costs:

Related Story An NVIDIA RTX GPU Couldn’t Be Used In A Rackmount Server Due To Non-Standard ATX Power Cables, So Modders Soldered Molex-Style Connectors To Make It Possible
  1. 100,000 GPUs x 24 hours x 90 days x $2.50 per GPU hourly cost = $540 million
  2. 100,000 GPUs x 24 hours x 120 days x $2.50 per GPU hourly cost = $720 million
  3. 100,000 GPUs x 24 hours x 90 days x $3.50 per GPU hourly cost = $756 million
  4. 100,000 GPUs x 24 hours x 120 days x $3.50 per GPU hourly cost = $1.008 billion

Do note that OpenAI conducted this training run most likely at its Stargate data center in Abilene. Of course, had the AI lab used one of CoreWeave's facilities, it likely would have incurred substantially higher costs, especially as CoreWeave sells its compute at around $10.50 per GPU hour.

What's more, NVIDIA's Jensen Huang publicly congratulated OpenAI just a few hours back, going on to announce a much heftier 400,000-GPU training run. Assuming this training run leverages NVIDIA's GB 200 NVL72 server racks, and where each rack consumes around 140kW of power, OpenAI will need ~1.2GW of power to complete this training run (5,555.55 racks x ~140kW x 1.5 PUE or Power Usage Effectiveness).

For the benefit of those who might not be aware, instead of requiring developers to furnish dedicated APIs for every application an AI agent needs to use, GPT-6 Astra is able to navigate any given software in a manner that is strikingly similar to how humans do so: by spawning agents that then work across browsers, spreadsheets, websites and desktop applications, producing finished documents and presentations, and implementing multistep workflows rather than merely telling a user how to complete them.

Basically, GPT-6 Astra's core reasoning engine utilizes a native multi-agent structure to solve complex problems. When faced with a complex task, the model's main orchestration agent can formulate a theory and deploy distinct sub-agents to test variations and validate results in parallel. This native delegation makes it highly resilient against "doom loops" - where a model gets stuck in repetitive error cycles - allowing it to autonomously debug its own code and adjust its strategy on the fly.

Even so, some remain critical as to whether OpenAI's GPT-6 Astra truly constitutes AGI. After all, Anthropic's Fable 5.1 remains ahead of Astra on SWE-bench Pro, Artificial Analysis Intelligence Index, Artificial Analysis Coding Agent Index, and ProofBench v1.1 (formal math proofs).

Rohail Saleem Photo

About the author: Writing is my one incontrovertible passion. Over the past six years, he has authored over 2,200 distinct articles on financial and tech-related topics, spanning nearly 1 million words. And he has been a member of Wcctech mobile team since 2025. As an alumnus of the University of Toronto, Rotman Commerce Program, I bring nuance, in-depth knowledge, and a unique perspective to every topic that I cover. When I'm not writing, I'm traveling the world, exploring hidden confectionaries and restaurants as an aspiring food connoisseur.

Follow Wccftech on Google to get more of our news coverage in your feeds.

Further Reading

A PNY graphics card with triple fans is displayed next to an NVIDIA chip labeled 'GB203-300-A1' with the text 'GB203' in large letters above them.

NVIDIA RTX 5070 Reportedly Gets GB205 Replaced By Bigger GB203 Die In A New Variant

A gameplay screenshot of 'Cyberpunk 2077' showing a mission prompt 'Go to Delamain HQ' in a street scene, alongside a system performance overlay listing 'NVIDIA GeForce RTX' GPUs at 0% usage for both GPU 0 and GPU 1.

User Demonstrates Offloading DLSS 5 Neural Rendering To The Second RTX 5060 Ti Using ReShade Add-On, Restoring Original Performance

An NBA 2K27 player wearing an Indiana Pacers jersey with '0' stands on a basketball court beside text stating, 'Guided by Artists, Powered by RTX' and 'Available Sept 3,' alongside the NVIDIA GeForce RTX logo.

NVIDIA DLSS 5 Cripples Most PCs in Modded Games, but DLSS 5 Autopilot Restores Playable Performance

A spiral galaxy appears in the center with the words 'GPT' on the left and 'Astra' on the right.

OpenAI’s GPT-6 Astra Spawns Armies Of Agents That Run On Local CPUs, Handing Intel And AMD A Demand Windfall

Read all comments on OpenAI’s ~$1 Billion GPT-6 Astra Heralds The Age Of AGI, According To NVIDIA’s Jensen Huang, Yet Anthropic’s Fable 5.1 Still Beats It On SWE-Bench Pro And Coding Benchmarks

Trending Stories

Popular Discussions

출처: Wccftech

작성자: Rohail Saleem

원문 게시일: 2026-09-07T14:53:05Z

이 글은 NewsAPI를 통해 자동으로 수집된 콘텐츠입니다. 원문은 위 링크를 통해 확인하세요.

AI ArtificialIntelligence OpenAI LLM Wccftech
← 목록으로