Editorial illustration for LWiAI Podcast #238: GPT 5.4 Mini, OpenAI Pivot, Mamba 3, Attention Residuals
GPT-5.4 Mini & OpenAI's Latest AI Model Breakthroughs
LWiAI Podcast #238: GPT 5.4 Mini, OpenAI Pivot, Mamba 3, Attention Residuals
The dust has barely settled on last week’s AI news cycle, and the landscape has already shifted again. OpenAI drops GPT‑5.4 Mini and Nano, faster, yes, but the price hike stings. Meanwhile, the company reportedly pivots hard toward business and productivity, leaving its consumer ambitions in the rearview.
Elsewhere, Mamba 3 promises a new architecture for efficiency, and attention residuals quietly rewrite how models learn. Mistral counters with a 128‑expert Small 4 model that punches far above its weight. NVIDIA opens its agent platform.
Meta delays its own AI. Episode 238 of LWiAI breaks down every announcement, every pivot, and every hidden signal. Buckle up.
Note from Andrey: this ep came out a week ago on RSS, but I was delayed posting it to youtube and therefore also Substack.
And so, the landscape shifts again. GPT-5.4 Mini arrives faster, sharper, and with a price tag that makes you wince. OpenAI pivots hard toward the boardroom, leaving the wilder edges of its ambition to the advisors who warned about them.
Mistral punches above its weight with 128 expert modules, while NVIDIA builds the scaffolding for a new generation of autonomous agents. DLSS 5 turns video games into a real-time generative filter, and Meta quietly delays its own AI rollout. The pattern is clear: the frontier is no longer just about capability.
It’s about control. Who builds the tools, who pays for them, and who decides what they’re allowed to see. Mamba 3 and Attention Residuals whisper that the architecture itself is still in flux, nothing is settled.
Not the models, not the markets, not the rules. The only constant is acceleration. The question isn’t whether you can keep up.
It’s whether you’re building something that matters while the ground moves beneath your feet.
Related Reading
Browse all OpenAI news →
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Related Topic OpenAI hires 630 ex-Meta staff as ChatGPT memory may turn data into ads
Business & Startups
Related Topic Sen. Warren urges Trump admin to expand AI credit after OpenAI letter
Business & Startups
From Archives Meta's structured prompting lifts LLM code review accuracy to 93%
From Archives Claude leak of 512,000+ lines reveals Tamagotchi-style pet, always‑on agent
From Archives OpenAI launches Codex plugin for Claude Code with local settings and Review Gate
Business & Startups
From Archives Disney pledged USD 1 billion equity stake in OpenAI and to become a major customer
Business & Startups
Common Questions Answered
What are the key specifications of OpenAI's new GPT-5.4 mini and nano models?
The new GPT-5.4 mini and nano models feature 400k-token context windows and promise faster inference with improved capabilities. However, they come with a significant pricing increase, with per-token rates up to four times higher than previous model generations.
How does Mistral's new Small 4 model differentiate itself in the AI landscape?
Mistral's Small 4 model stands out by incorporating 128 expert modules, allowing it to perform beyond expectations for its size. This approach suggests a novel architectural strategy that potentially challenges larger, more established AI models.
What notable AI developments were discussed in the LWiAI Podcast #238 episode?
The podcast covered multiple AI developments, including OpenAI's GPT-5.4 mini and nano releases, Mistral's Small 4 model with 128 expert modules, and Meta's 'My Computer' project that transforms Macs into AI agents. The episode was recorded on March 18, 2026, and later published on YouTube and Substack.
