ChatGPT-5 is here - Weekly Wrap - Lighthouse AI Newsletter
A lot happened this week in the world AI and that was before yesterdays launch of ChatGPT-5 so a lot to cover in this weeks newsletter – lets dive in.
AI News (Macro, Economic & Geopolitical Events)
OpenAI Hits $12 Billion in Annualized Revenue, Breaks 700 Million ChatGPT Weekly Active Users
OpenAI roughly doubled its revenue in the first seven months of 2025, reaching $12 billion in annualized revenue and generating approximately $1 billion per month, according to The Information. ChatGPT is on track to hit 700 million weekly active users this week, up from 500 million in March, marking a fourfold year-over-year surge. The company now has 5 million paying business users, up from 3 million in June, while its annual recurring revenue has reached $13 billion as it prepares for GPT-5’s launch. Link: https://www.theinformation.com/articles/openai-hits-12-billion-annualized-revenue-breaks-700-million-chatgpt-weekly-active-users
Anthropic revokes OpenAI's API access
Anthropic has terminated OpenAI’s API access to its Claude models, alleging that OpenAI violated its terms of service. The justification provided was that OpenAI was utilizing Anthropic’s coding tools for evaluation and benchmarking ahead of the anticipated launch of GPT-5. This action represents a significant escalation in competitive tensions between the two leading AI labs, with an OpenAI spokesperson describing its usage as “industry standard” while expressing disappointment with the decision. Link:https://www.reuters.com/technology/anthropic-restricts-openais-access-its-ai-models-2025-08-01/
Runway, in Talks for Multibillion-Dollar Fundraises as Revenue Picks Up
Runway, a startup that has cut deals with Lionsgate and AMC Networks, is in talks with investors to raise around $500 million at a valuation of at least $5 billion before the investment, representing a 67% increase from the six-year-old company’s last round a year ago. The AI video generation startup’s previous funding was $308 million at a $3 billion valuation in April 2025. The new funding discussions come as Runway and competitor Luma seek significant capital amid growing demand for AI video generation tools from Hollywood studios and content creators. Link: https://www.theinformation.com/articles/runway-luma-talks-multibillion-dollar-fundraises-revenue-picks
Luma AI targets up to $1.1 billion funding at $3.2 billion valuation amid AI video boom
Luma AI, known for its Dream Machine tool that generates hyper-realistic video clips, is in discussions to raise up to $1.1 billion at a $3.2 billion valuation, reflecting strong investor optimism despite modest revenue figures. The AI video startup, which previously raised $90 million in December 2024 led by Amazon, is exploring expansions beyond pure video generation into robotics and enterprise applications. The company reported an $8 million annualized run rate as of August 2024, with growth driven by premium subscriptions and partnerships as it competes with rivals like Runway in the rapidly evolving AI video generation market. Link: https://www.webpronews.com/runway-luma-seek-1-6b-funding-at-sky-high-valuations-despite-low-revenues/
IBM announces plans to acquire DataStax for enterprise AI data
IBM announced plans to acquire DataStax, aiming to help unlock enterprise data for AI applications while expanding its NoSQL database offerings. The acquisition is part of IBM’s strategy to strengthen its AI infrastructure capabilities for enterprise clients. Link:https://newsroom.ibm.com/2025-08-06-IBM-Announces-Intent-to-Acquire-DataStax-to-Help-Clients-Unlock-Enterprise-Data-for-AI
Tech AI News & Developments
Google releases Gemini 2.5 Deep Think
Google released Gemini 2.5 Deep Think, its first publicly available multi-agent model that does “parallel thinking” to help researchers, scientists, and academics tackle complex problems. The model spawns multiple agents to explore possible solutions in parallel and then decides the best answer from them. Link:https://blog.google/technology/ai/google-gemini-2-5-deep-think/
Anthropic releases Claude Opus 4.1 with enhanced coding and reasoning capabilities
Anthropic released Claude Opus 4.1 as an upgrade to Claude Opus 4 on agentic tasks, real-world coding, and reasoning. The model advances state-of-the-art coding performance to 74.5% on SWE-bench Verified and improves Claude’s in-depth research and data analysis skills, especially around detail tracking and agentic search. Link: https://www.anthropic.com/news/claude-opus-4-1
Google NotebookLM rolls out Video Overviews and upgraded Studio panel
Google announced that it’s rolling out Video Overviews to NotebookLM, its AI-based note-taking and research assistant. The feature allows users to turn dense multimedia including raw notes, PDFs, and images into digestible visual presentations with AI-narrated explainer videos. The update also includes a redesigned Studio panel that lets users create and store multiple studio outputs of the same type in a single notebook, with new tiles for creating Audio Overviews, Video Overviews, Mind Maps, and Reports with a single click. Link: https://blog.google/products/notebooklm/notebooklm-video-overviews-studio-panel/
OpenAI releases a free GPT model that can run on your laptop
OpenAI is releasing a new open-weight model dubbed GPT-OSS that can be downloaded for free, be customized, and even run on a laptop. This is OpenAI’s first open-weight model in over six years, coming in 120-billion-parameter and 20-billion-parameter versions under the Apache 2.0 license.Link: https://openai.com/index/gpt-oss/
Ideogram Introduces Character Consistency Model
Ideogram released Character, an AI model that uses one reference image to generate consistent characters across styles, expressions, scenes, and lighting. Free for all web and iOS users, enhancing creative workflows. Link: https://ideogram.ai/blog/character
Perplexity launches cheaper Deep Research tool to compete with OpenAI
Perplexity introduced its Deep Research tool powered by Deepseek-R1, offering autonomous research capabilities at significantly lower pricing than competitors. While OpenAI charges $200/month for 100 queries, Perplexity offers 500 queries per day for $20/month. Link:https://blog.perplexity.ai/blog/introducing-deep-research
Xbox announces "Copilot for Gaming" AI assistant for gamers
Xbox introduced “Copilot for Gaming,” a voice-activated AI companion that offers real-time gaming tips, advice, and social features. The assistant can notify users when friends are online and adapts to both multiplayer and solo gaming styles. Link: https://news.xbox.com/en-us/2025/08/06/introducing-copilot-for-gaming/
Alibaba releases open-source AI video suite that outperforms Sora
Alibaba’s Tongyi Lab released Wan2.1, an open-source suite of powerful video generation models that outperform state-of-the-art models including Sora on key benchmarks. The suite supports text-to-video, image-to-video, and video-to-audio generation at 2.5x the speed of competitors. Link: https://www.alizila.com/alibaba-releases-open-source-ai-video-suite-wan-2-1/
ElevenLabs launches an AI music generator that it says is cleared for commercial uses
ElevenLabs has launched its latest AI audio product with Eleven Music, claiming it “is cleared for nearly all commercial uses, from film and television to podcasts and social media videos, and from advertisements to gaming.” Caveat I am no musician, but I know what I like, I had a play with ElevenLabs Music, more trying to create some background music, but it produced a full song – I have heard worse at Eurovision. Link: https://elevenlabs.io/blog/elevenlabs-music-release/
xAI launches 'Grok Imagine' video generator
xAI has officially released Grok Imagine, its new AI image and video generator, making it available to all SuperGrok and Premium+ subscribers on the X iOS app. The tool enables users to convert text prompts or existing images into 15-second videos, complete with native audio, in a fraction of the time taken by competing models. This launch positions xAI to compete in the consumer AI space by prioritizing speed and tight integration with the X social media platform. Link: https://x.ai/blog/grok-imagine
Runway’s Aleph for AI-Powered Video Editing
Runway has introduced Aleph, a new “in-context” video model that allows for sophisticated editing and transformation of existing video footage through text prompts. The model can perform tasks such as generating new camera angles, applying stylistic transfers, and adding or removing objects within a scene. This technology moves beyond simple video generation to offer granular post-production control, targeting professional and enterprise users. Link: https://runwayml.com/blog/introducing-aleph/
Tools in Focus - ChatGPT-5: The AI PhD That Fits in Your Pocket
OpenAI’s fifth‑generation model is more than a routine upgrade; it jumps from “clever student” to pocket‑sized panel of specialists. The new version promises to be smarter, faster, and more intuitive than any of its predecessors. While around 700 million people currently use ChatGPT weekly, the company anticipates that GPT-5 will redefine how individuals and businesses interact with artificial intelligence.
ChatGPT 5 was released yesterday PM and the plan is for it to quickly replace all the previous models and have the entire user base switch to GPT-5. At the time of writing this despite repeated refreshed we hadn’t got access to 5 so now hands on experience yet. Here is a summary of what was said during the 1:20 minute launch.
Expert reasoning without the wait GPT‑5 blends the speed of “fast” models with the deliberative thinking of the reasoning line. It pauses just long enough to consider a question before delivering answers that outperform experts on multi‑modal and mathematical benchmarks, whilst sharply reducing hallucinations and deceptive slip‑ups.
Software on demand Ask for an application and GPT‑5 will write, test and iterate an entire codebase, front‑end, back‑end and infrastructure together, explaining every step. It tops Swe‑bench, MATH and tool‑calling evaluations, so for the first time an AI can behave like a senior engineer rather than an intern.
For developers, GPT-5 sets a new standard in coding. It can write entire computer programs from scratch, a concept OpenAI calls “software on demand”. This is showcased by its ability to build interactive web applications and detailed dashboards from simple, concise prompts. The model also demonstrates superior performance on complex coding benchmarks like SWE-bench.
Personalised voice and memory New features extend to the user experience. Personalisation has been enhanced, allowing users to select different chatbot personalities, from supportive to sarcastic. The voice model is now more natural and features a “Study and Learn Mode” for guided learning. In a significant move, ChatGPT can now connect to a user’s Gmail and Google Calendar to assist with scheduling and daily planning.
Healthcare‑grade accuracy On the physician‑designed HealthBench, GPT‑5 is the most reliable model yet for clinical reasoning, translating jargon‑heavy notes into plain language, flagging missing information and outlining treatment options.
Enterprise strength A 400k‑token context window can absorb code repositories or legal bundles. New API tricks such as custom tool definitions, regex‑bounded outputs and verbosity controls arrive with a three‑tier family (GPT‑5, GPT‑5 Mini, GPT‑5 Nano), letting teams balance power with cost.
Safety baked in, safety protocols have also been overhauled. GPT-5 introduces “safe completions,” which aim to provide helpful responses within safety constraints, reducing outright refusals and guiding users towards safer information.
GPT-5 is rolling out to all users, including those on the free tier for the first time. The API will also feature GPT-5, alongside the more affordable Mini and Nano models, making this powerful technology accessible to a broader audience