GPT-4 news
The latest GPT-4 news, : NVIDIA Switchyard routes 93% of tasks to cheap models but tool definitions break lightweight models.
GPT-4 is a large language model used in diagnostic trials where it outperforms physicians on specific questions, though safety concerns remain. Recent coverage highlights OpenAI's strategic shift, with 80 to 90 percent of research efforts now targeting GPT 7 and future generations rather than current models.
16 stories mentioning GPT-4 since 15 September 2026, newest first, each with its sources and discussion. Follow to see new ones on your front page.
-
NVIDIA Switchyard routes 93% of tasks to cheap models but tool definitions break lightweight models
NVIDIA's open-source Switchyard library routes tasks between models based on complexity, and a LangChain benchmark on 145 tasks showed 93% went to lightweight models with a 6% performance drop and 74% cost reduction.…
1 sourcenote.com -
GEPI report finds AI systems amplify gender bias across hiring, healthcare and lending
The Gender Equity Policy Institute released a report titled "How AI Fails Women: When Bias Becomes Amplified," authored by senior fellow Myriam Shiran, PhD. The report synthesizes existing academic and gray literature…
1 sourceUnite.AI -
Co-Existence released, uses Anthropic Opus 5.5 to chart 30 months of AI progress
The author’s new book Co‑Existence launched today, looking back at AI development since the early‑2024 release of Co‑Intelligence. Using Anthropic’s Opus 5.5 in Claude Projects, the author let a single “orchestrator”…
1 sourceOne Useful Thing -
Opinion: AI scales human bias instead of eliminating it
AI systems inherit the judgments and blind spots embedded in the data they are trained on, so when they are used for hiring, healthcare or employee evaluation they can amplify existing discrimination at a speed no…
1 sourceUnite.AI -
Researchers test open-weight Hunyuan-MT-7B for stylistic translation
A new paper on arXiv explores whether an open-weight model can replicate GPT-4’s stylistic translation capabilities. The study compares three variants of Hunyuan-MT-7B—base, LoRA fine-tuned, and GRPO-steered—against…
1 source primary sourcearXiv cs.CL -
DHS bans DeepSeek AI from civilian agencies, extends to third-party products developed using its models
The Department of Homeland Security filed an exclusion order on Monday removing all DeepSeek and DeepSeek-affiliate products from federal civilian executive branch agencies — roughly 430 departments — and banning new…
40 sources HN 6The Decodernytimes.comInterconnectsTechCrunch AI +33 more -
Dyna Robotics deploys Dyna-2.1 robots folding napkins at Din Tai Fung with 99.4% success rate
Dyna Robotics co-founder and chief scientist Jason Ma detailed the company's Dyna-2 world-action model and its physical deployment Dyna-2.1 on the Latent Space podcast. The model was pre-trained on over one million…
1 sourceLatent Space -
Kevin Roose publishes The AGI Chronicles after 150 interviews on AI race
New York Times columnist Kevin Roose has published The AGI Chronicles, a book chronicling the recent history of generative AI through the stories of OpenAI, Anthropic, and Google. Roose conducted more than 150…
1 sourceWired AI -
AI outperforms physicians on some diagnostic tasks, but safety concerns remain
Recent trials show large language models can beat doctors on specific diagnostic questions. In a 2024 randomized study, physicians using GPT‑4 scored 76 % versus 74 % with conventional resources, a gap that was not…
1 sourceUnite.AI -
MIT, Google and Northeastern unveil InstructMesh for editing AI‑generated 3D models
InstructMesh is a new interface that lets users generate, edit, and 3D‑print objects with AI assistance. Developed by MIT’s CSAIL, Google, Microsoft’s TRELLIS system and Northeastern University, the tool pairs the…
1 source primary sourceMIT News on AI -
AI Foundations publishes 2026 guide comparing ChatGPT Free, Plus and Pro tiers
AI Foundations released a guide that walks through ChatGPT's three subscription tiers as of 2026: Free, Plus and Pro. The Free plan offers basic functionality for casual users. The Plus plan provides faster response…
1 sourcegeeky-gadgets.com -
OpenAI says 80 to 90 percent of research targets GPT 7 and beyond
Boris Power, OpenAI's Head of Applied Research, stated that 80 to 90 percent of the company's research efforts are directed toward GPT 7, GPT 8, and future generations. Power explained that this focus exists because…
1 sourceThe Decoder -
Russian AI gateways charge 2.4-3.6x list price for token access
Russian developers face significant price markups for AI tokens due to sanctions that block direct payment to Western providers like OpenAI and Anthropic. A report details how local gateways, such as AITUNNEL and…
1 source HN 41infertrail.com -
Claude cannot forecast crypto or stock prices, benchmark shows humans outperform
Anthropic’s Claude model does not have live market data and cannot predict cryptocurrency or stock prices. A peer‑reviewed benchmark presented at ICLR 2025 found that expert human forecasters beat the best large…
2 sourcesnote.comhackernoon.com -
Microsoft exec calls AI scraping the largest theft of labor, internal filings show
Unsealed filings in the New York Times copyright lawsuit reveal internal Microsoft and OpenAI documents in which Director of Applied Science Brent Hecht described AI scraping as “the largest theft of labor in human…
7 sources HN 676The Verge AITom's Hardwareyahoo.comtechspot.com +3 more -
The 6 best AI writing generators for 2026
Zapier’s latest roundup evaluates six AI‑powered writing tools that have survived the rapid churn of the past few years. The list includes Jasper, Anyword, Writer, Writesonic, Rytr and Sudowrite, each judged on…
1 sourceZapier Blog
Questions about GPT-4
What is the latest GPT-4 news?
As of 21 October 2026: NVIDIA Switchyard routes 93% of tasks to cheap models but tool definitions break lightweight models. NVIDIA's open-source Switchyard library routes tasks between models based on complexity, and a LangChain benchmark on 145 tasks showed 93% went to lightweight models with a 6% performance drop and 74% cost reduction.
What is GPT-4?
GPT-4 is a large language model used in diagnostic trials where it outperforms physicians on specific questions, though safety concerns remain. Recent coverage highlights OpenAI's strategic shift, with 80 to 90 percent of research efforts now targeting GPT 7 and future generations rather than current models.