Updates and recent posts about GPT-5.4..

Posts
Description

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

Researchers Fine-Tune LLM for Reasoning with Only 1,000 Examples

Meet the"Wait" token trick—a clever nudge that sharpens a model's reasoning. It mirrors OpenAI's o1-preview magic using only 1,000 examples. And guess what? Not a speck of reinforcement learning in sight... read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

3: Think Deeper, Act Faster

Qwen3sets itself apart with its dazzlingHybrid modes. Flip between deep thought and rapid-fire replies. A magician capable of juggling complexity and speed. Themassive 235B modelthrows elbows with the high rollers in AI town. Meanwhile, the nimble30B MoE variantdazzles with its frugality, flexing st.. read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

Foundation Model for Personalized Recommendation

Netflixhas given its recommender system a makeover with a foundation model similar toLLMs. The goal? Turbocharge efficiency and scalability by making member preferences the star of the show. They turned user interactions into tokens, kind of like BPE in NLP, and employedsparse attentionto zero in on.. read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

Cloudflare CEO: AI is killing the business model of the web

Google's dominance in search is fading due to AI, leading to a decline in traffic for content creators, threatening the web's sustainability... read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

Amid DeepSeek Threat Sundar Pichai-Led Google Ups AI-Generated Coding By Over 30% To Boost Productivity

Google now churns out more than 55% of its code with AI, a big leap from last year's 25%.Meanwhile, CEO Sundar Pichai plays it cool, warning we're still in the AI toddler phase. But they're not just tinkering. Google's diving headfirst into AI Modes with Search, aiming to flip the script for a billi.. read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

AI use damages professional reputation, study suggests

Duke University reveals a startling twist: AI tools like ChatGPT don't just supercharge work; they also slap users with unfair labels.Lazy. Replaceable. These biases stick to everyone, demographics be damned. Even when productivity soars, fellow workers and bosses often question AI users' competence.. read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

OpenAI plans to release a new 'open' AI language model in the coming months

OpenAI's having a change of heart. Picture a reluctant flipper resting on the high-dive, finally plunging into open waters. They're ready to unleash an“open” language model, thanks to pressure from competitors likeDeepSeekandMetawho have been living the open-source dream. CEO Sam Altman has conceded.. read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

Coding emerges as generative AI’s breakout star

AI coding tools are revolutionizing software development, with many developers already using them for efficiency gains. OpenAI's latest model ranks in the top competitive coders percentile, showing rapid progress in reasoning abilities. AI coding tools are set to support huge context windows, potent.. read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

Alibaba’s ‘ZeroSearch’ lets AI learn to google itself — slashing training costs by 88 percent

Alibaba researchers developed ZeroSearch to train large language models (LLMs) to search for information without using real search engines, reducing costs by up to 88%. ZeroSearch outperformed Google in experiments, demonstrating the potential for AI systems to simulate search and reduce reliance on.. read more

Link

@faun shared a link, 1 year, 1 month ago

FAUN.dev()

‘AI is already eating its own’: Prompt engineering is quickly going extinct

AI prompt engineering has vanished as a standalone job, absorbed into general AI roles. New AI roles demand deeper technical expertise and are reshaping the job market quickly... read more

GPT-5.4 is OpenAI’s latest frontier AI model designed to perform complex professional and technical work more reliably. It combines advances in reasoning, coding, tool use, and long-context understanding into a single system capable of handling multi-step workflows across software environments. The model builds on earlier GPT-5 releases while integrating the strong coding capabilities previously introduced with GPT-5.3-Codex.

One of the defining features of GPT-5.4 is its ability to operate as part of agent-style workflows. The model can interact with tools, APIs, and external systems to complete tasks that extend beyond simple text generation. It also introduces native computer-use capabilities, allowing AI agents to operate applications using keyboard and mouse commands, screenshots, and browser automation frameworks such as Playwright.

GPT-5.4 supports context windows of up to one million tokens, enabling it to process and reason over very large documents, long conversations, or complex project contexts. This makes it suitable for tasks such as analyzing codebases, generating technical documentation, working with large spreadsheets, or coordinating long-running workflows. The model also introduces a feature called tool search, which allows it to dynamically retrieve tool definitions only when needed. This reduces token usage and makes it more efficient to work with large ecosystems of tools, including environments with dozens of APIs or MCP servers.

In addition to improved reasoning and automation capabilities, GPT-5.4 focuses on real-world productivity tasks. It performs better at generating and editing spreadsheets, presentations, and documents, and it is designed to maintain stronger context across longer reasoning processes. The model also improves factual accuracy and reduces hallucinations compared with previous versions.

GPT-5.4 is available across OpenAI’s ecosystem, including ChatGPT, the OpenAI API, and Codex. A higher-performance variant, GPT-5.4 Pro, is also available for users and developers who require maximum performance for complex tasks such as advanced research, large-scale automation, and demanding engineering workflows. Together, these capabilities position GPT-5.4 as a model aimed not just at conversation, but at executing real work across software systems.