Tech News
← Home  ·  All topics

Gpt

251 GoKawiil briefs on this topic · RSS for gpt

Anthropic Publishes Early Framework for Reverse-Engineering Transformer Circuits

Anthropic researchers introduce a mathematical approach to mechanistic interpretability, aiming to reverse-engineer the internal computations of transformer language models. Their initial study focuses on small transformers with two layers or fewer that use only attention blocks, deliberately simpler than models like GPT-3, in order to identify basic patterns before tackling larger systems.

Developer finds GPT-6 'Astra' burns billions of tokens with no usable code output

A software engineer tested OpenAI's GPT-6 Astra model by setting up an autonomous 'software factory' where the model managed its own workflow, context and subagents to build a Python variant with virtual threads and lexical scoping. After roughly 35 hours and about 4 billion tokens consumed, the effort produced nothing usable and taught the author little about how to actually apply the model to real software engineering work.

OpenAI brings GPT-Live-1 voice model to its API for developers

OpenAI has opened API access to GPT-Live-1, a voice model that can listen and speak simultaneously and hand off complex reasoning or tool use to text models like GPT-6 Astra. The model, already used in ChatGPT, lets developers customize tone and pacing, handle background noise, and support full-duplex phone-based voice agents. Early testing by language-learning company Speak found it cut conversational interruptions by nearly 80% compared with older turn-based voice systems.

OpenAI Launches ChatGPT for Financial Services With Morgan Stanley, Evercore

OpenAI has introduced ChatGPT for Financial Services, a Wall Street-tailored version of ChatGPT Work built with Morgan Stanley and Evercore, powered by its new GPT-6 Astra model. The tool can research potential M&A targets, pull data from industry-standard financial sources, and automatically generate formatted PowerPoint pitch decks matching a bank's style guide.

OpenAI launches ChatGPT for Financial Services with Morgan Stanley, Evercore

OpenAI has released ChatGPT for Financial Services, a specialized version of its enterprise product built with input from Morgan Stanley and Evercore. Powered by its newest model, GPT-6 Astra, the tool can research companies, analyze financial data and automatically build the pitch decks that junior investment bankers typically produce by hand.

Study: Qwen3.8 answers shift sharply toward GPT-5.5 Pro under reasoning-prefill test

A follow-up experiment tested four open models—DeepSeek V4 Flash, Inkling, Kimi K3, and Qwen3.8 A95B—by inserting the first 1% of GPT-5.5 Pro's reasoning trace into each model's own reasoning channel before letting it generate answers freely. Researchers then measured how much of GPT-5.5 Pro's visible answer text overlapped with each model's output. Qwen3.8 showed the largest jump, with overlap rising from 33.92% unprefilled to 54.50% with the GPT-5.5 Pro prefill, a 20.58 percentage-point increase, while other models showed much smaller shifts.

OpenAI's GPT-6 Astra sparks debate over hidden reasoning and looped transformer design

Analysts examining OpenAI's newly released GPT-6 Astra note it outperforms its GPT-5.6 predecessor across writing, math, coding, and especially 3D rendering tasks, jumping from 7.8% to 99.9% on the ARC-AGI-3 benchmark. Speculation has emerged that Astra uses a looped or recurrent-depth transformer architecture and may be concealing parts of its chain-of-thought reasoning process from users.

OpenAI's GPT-6 Astra sparks AGI debate as chief scientist calls it an 'Alien Mind'

OpenAI released GPT-6 Astra, which the company describes as its safest and most capable model yet, trained on roughly 100,000 Nvidia Blackwell GPUs. Nvidia's Jensen Huang publicly declared that AGI has been achieved, while OpenAI chief scientist Jakub Pachocki offered a more cautious framing, likening the model to an 'alien mind' that only simulates aspects of human behavior rather than replicating it. Independent benchmarks place Astra roughly on par with rival Fable 5.1, though reportedly cheaper to run per task.

Nvidia's Jensen Huang again calls AI 'AGI' while touting Blackwell GPU sales

Jensen Huang has for the second time this year proclaimed that artificial general intelligence has been achieved, this time tying his claim to OpenAI's newly launched GPT-6 Astra model. He highlighted that Astra was trained on more than 100,000 Nvidia Blackwell chips, with 400,000 more coming online soon, even as independent benchmarks show Astra trailing rival Claude in some tasks.

OpenAI upgrades image tool with ChatGPT Images 2.5 release

OpenAI has launched ChatGPT Images 2.5, an update to its April image generation model that promises sharper detail, more natural lighting, and better consistency when editing across multiple prompts. The company says generation latency has dropped by up to 50% compared with the original Images 2 model, and it's adding a sketch-to-image feature plus templates and prompt-sharing tools. Developers also get two new API models, GPT-Image-2.5 Flare and Sunburst, aimed at different levels of editing precision and speed.

OpenAI launches ChatGPT Images 2.5, cutting generation latency by 50%

OpenAI released Images 2.5, an upgraded image model powering ChatGPT and the GPT-Image API, offering sharper detail, better subject preservation from reference photos, and more reliable multi-turn editing. The update ships alongside new ChatGPT features like Sketch, templates, and image comments, plus two developer-facing API models, GPT-Image-2.5 Flare and Sunburst.

OpenAI system produces proof of finite-time singularity in Navier–Stokes equations

OpenAI says an internal AI model, more capable than GPT-6 Astra, generated a solution to part of the Navier–Stokes existence and smoothness problem, one of the Millennium Prize Problems. The proof demonstrates that three-dimensional incompressible fluid dynamics governed by these equations can develop a singularity—unbounded speed growth—in finite time, even from smooth initial conditions. OpenAI released both a written explanation of the proof and a formal verification built in the Lean proof assistant.