OpenAI's new GPT-5 roadmap

PLUS: Adobe launches IP-safe AI video generator

Author

Rowan Cheung
February 13, 2025

Good morning, AI enthusiasts. Sam Altman just dropped OpenAI’s new model roadmap and scrapped the standalone release of o3 — but something much bigger is taking its place.

With a new plan to unify the company’s tools and models under an upcoming GPT-5 umbrella, could the days of the fragmented AI ecosystem finally end?

Reminder: Our next workshop is today at 4 PM EST, where you’ll learn how to build your own AI-powered personal assistant to streamline your daily tasks! RSVP here.

In today’s AI rundown:

  • Sam Altman details GPT-5 roadmap, scraps o3 launch

  • Adobe launches IP-safe AI video generator

  • Repurpose video transcripts across platforms

  • OpenAI expands Model Spec with focus on intellectual freedom

  • 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

OPENAI

Image source: Sam Altman (@sama) on X

The Rundown: CEO Sam Altman just posted a roadmap update for OpenAI’s upcoming models, canceling its anticipated o3 release in favor of an integrated GPT-5 system that will unify and simplify the companies’ AI offerings.

The details:

  • GPT-5 will incorporate o3's capabilities alongside other OpenAI tech, with a tiered system offering different intelligence levels for free, Plus, and Pro users.

  • Before GPT-5, the company will release GPT-4.5 (codenamed "Orion") as its final non-chain-of-thought model in the coming weeks.

  • Free users will have unlimited access to GPT-5 at "standard intelligence," while paid tiers unlock higher performance and advanced tools.

  • Altman provided an ETA of ‘weeks’ for 4.5 and ‘months’ for 5, with o3 no longer being released as a standalone model.

Why it matters: The model landscape (and naming conventions) have become increasingly difficult to follow, making for a poor user experience. This new unified structure sounds like GPT-5 will automatically select the best models and tools for a specific task, potentially reshaping how intelligence is packaged and delivered to users.

TOGETHER WITH UNSTRUCTURED

The Rundown: We’re all building our GenAI stacks on the fly. Join Unstructured & Elastic for a live session to accelerate your learning curve and learn the latest techniques for fueling your apps and pipelines with all your document data.

Join the live webinar on Jan. 22 to learn how to:

  • Build streamlined ETL pipelines for your unstructured GenAI data

  • Balance good/fast/affordable parsing for your most complex documents

  • Parse and search everything once your data is within Elastic

  • Unlock internal business data with a live code demo of actionable steps

Click here to register and receive the recording, even if you can’t attend live.

ADOBE

Image source: Adobe

The Rundown: Adobe just released its Firefly Video Model in public beta, positioning it as the industry's ‘first commercially safe AI video generation tool’ — along with new pricing tiers starting at $9.99 for its expanded Firefly creative application suite.

The details:

  • The new system can generate 1080p video clips from text or images, with precise camera control, shot framing, and motion graphics capabilities.

  • The model is trained on licensed Adobe Stock and public domain content, and the company emphasizes that it will never be trained on customer generations.

  • Adobe is launching two new subscription tiers: Standard ($9.99/month for 20 videos) and Pro ($29.99/month for 70 videos).

  • Other upgrades include Translate and Lip Sync for audio, Scene to Image for 3D structure references, and broader integrations with other Adobe platforms.

Why it matters: There is no shortage of competition for the AI creative suite, but few companies have focused directly on being commercially safe — a differentiator that could make Firefly the go-to choice for brands and agencies that need legally safe AI-generated content at scale.

AI TRAINING

The Rundown: Claude can transform your video transcripts into multiple content pieces while maintaining your unique voice and style.

Step-by-step:

  1. Gather your transcript and 2-3 samples of your best content.

  2. Train Claude with your detailed style-matching prompt by instructing it to analyze your examples.

  3. Request to transform your transcript into focused, voice-matched content for one or multiple social media platforms.

Pro tip: Save time by creating a custom Claude Project trained specifically on your voice and style. The Rundown University member can access the top prompts here.

PRESENTED BY INNOVATING WITH AI

The Rundown: Innovating with AI’s new program, AI Consultancy Project, equips AI enthusiasts with all the resources to capitalize on the rapidly growing AI consulting market – which is set to 8x to $54.7B by 2032.

The program offers:

  • Tools and framework to find clients and deliver top-notch services

  • A 6-month roadmap to build a 6-figure AI consulting business

  • Student landing their first AI client in as little as 3 days

OPENAI

Image source: OpenAI

The Rundown: OpenAI just released an expanded Model Spec that outlines how models should handle controversial topics, emphasizing truth, user customization and intellectual freedom while maintaining key safety guardrails.

The details:

  • The 63-page specification introduces a "chain of command" where platform rules precede developer and user preferences.

  • After feedback requesting a "grown-up mode," OpenAI is exploring ways to allow types of adult content while maintaining strict bans on harmful material.

  • The company is combatting ‘AI sycophancy’ by training models to give honest feedback instead of empty praise and avoiding agenda-seeking responses.

  • The Model Spec is released under a CC0 license, allowing other AI companies to adopt and modify these guidelines for their own systems.

Why it matters: Another common critique of OpenAI is the frequent restrictions around controversial topics or conversations. With open source and unfiltered models like Grok gaining in popularity, it appears ChatGPT is going down the less restrictive, ‘truth-seeking’ route — a move even Elon Musk should ideologically approve of.

QUICK HITS

  • ⚡Deepgram - Introducing Nova-3, the first voice AI model with real-time multilingual speech-to-text*

  • 🧠 Perplexity Sonar - New upgraded model with 10x faster response speed

  • 🧮 DeepScaleR - Open-source 1.5B model with exceptional math reasoning

  • 🧊 3D AI Studio V3 - Generate, organize, and edit 3D models

    *Sponsored Listing

Apple analyst Ming-Chi Kuo revealed that Apple is exploring humanoid and non-humanoid robots for its smart home ecosystem, though mass production isn't expected before 2028.

Sam Altman said that OpenAI is planning to extend access to its Deep Research tool to all ChatGPT tiers, with 2 uses per month for free users and 10 for Plus users to start.

Thomson Reuters secured a landmark AI copyright legal victory, with a judge ruling that Ross Intelligence’s use of copyrighted content for AI training constituted infringement.

Midjourney founder David Holz teased that the company has two hardware projects currently in development, with ‘one that goes on you’ and ‘one that you go inside of.’

AI infrastructure startup fal secured a $49M Series B to expand its video-focused generative media platform, which already processes over 100M daily inference requests for enterprise customers including Quora and Canva.

Glean launched Glean Agents, a new platform allowing enterprises to build and deploy custom assistants with access to company and internet data.

COMMUNITY

Join our workshop today at 4 PM EST to learn how to build your own AI-powered personal assistant to streamline your daily tasks with Dr. Alvaro Cintas, The Rundown’s AI professor.

RSVP here. Not a member yet? Join The Rundown University on a 14-day free trial.

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team

Reply

or to participate.