|
In this newsletter:
- OpenAI launches smarter AI models at a fraction of the cost
- Grok 4.5 delivers top AI performance at half the cost
- ChatGPT voice now listens while you talk without interrupting
Plus, you’ll find new AI tools and this week’s top AI news headlines!
🚀 OpenAI Launches GPT-5.6 With Improved Intelligence and Better Performance
OpenAI released GPT-5.6 in three tiers - Sol, Terra, and Luna - available now across ChatGPT, Codex, and the OpenAI API, delivering stronger performance across coding, knowledge work, cybersecurity, and science while using fewer tokens than previous models.
Three model tiers:
- Sol: flagship model, $5 input/$30 output per million tokens
- Terra: balanced everyday model, $2.50 input/$15 output per million tokens
- Luna: fastest and most affordable, $1 input/$6 output per million tokens
What's new:
- Available now across ChatGPT, Codex, and the OpenAI API
- Plus, Pro, Business, and Enterprise users access Sol through medium and higher effort settings
- New "ultra" setting coordinates four parallel agents by default for faster results on demanding tasks
- New "max" setting gives model more time to reason, explore alternatives, and revise approach
- Programmatic Tool Calling lets the model write and run lightweight programs that coordinate tools and process results without routing everything back through the model
- New ChatGPT Work tool for documents, spreadsheets, and presentations running on desktop, web, and mobile
- Creates fully editable presentations from scratch with layouts, typography, and hierarchy
- Follows design system conventions from reference decks, including Slide Master rules
- Stronger computer use: inspects and refines rendered results, not just generates code
- Improved cache system with explicit cache breakpoints and 30-minute minimum cache life
Performance vs competitors:
- Agents' Last Exam professional workflows: Sol 52.7% vs Claude Fable 5 (40.5%), Opus 4.8 (45.2%)
- Coding Agent Index: Sol 80, above Fable 5 (77.2%), using less than half output tokens at one-third the cost
- OSWorld 2.0 computer use: Sol 62.6% vs Opus 4.8 (54.8%) using 85% fewer output tokens
- BrowseComp web search: Sol 90.4%, Sol Ultra 92.2%
- Terra and Luna both outperform Fable 5 at roughly one-sixteenth the estimated cost
- Lovable: 25% fewer steps and 35-48% fewer tool calls than prior model, 15% fewer stuck runs
Cybersecurity:
- Strongest cybersecurity model yet, blocks roughly ten times more potentially harmful activity than GPT-5.5
- ExploitBench: 73.5% vs GPT-5.5's 47.9%
- Supports secure code review, patching, threat modeling, and blue teaming
- Individual users must enable hardware-backed passkeys by September 1 to retain access to cyber-capable models
- Trusted Access for Cyber program available for verified defenders needing broader access
Science:
- GeneBench Pro genomics: Sol 28.7% vs GPT-5.5's 12%
- LifeSciBench: Sol 59.9% vs GPT-5.5's 50.4%
- Strong gains in biology, life science research workflows, and chemistry
Why it matters:
GPT-5.6 makes the strongest case yet that intelligence and cost efficiency don't have to trade off against each other - Sol beats or matches Fable 5 on most benchmarks while using far fewer tokens, and Luna delivers GPT-5.5-level results at a fraction of the price.
For businesses watching their AI spend, the three-tier structure gives a practical answer to the question of which model to use for which tasks, rather than forcing everyone onto the most expensive option just to get reliable results.
👀 Read more about ChatGPT's new models!
⚡Grok 4.5 Delivers Top AI Performance at Half the Cost
SpaceXAI released Grok 4.5, positioning it as a capable alternative to Anthropic's Opus models at significantly lower cost, with twice the token efficiency of leading competitors.
What's new:
- Available now to the public after beta testing
- Handles coding, app-building, office work, research, writing, and routine knowledge work
- Twice the token efficiency of other leading models according to SpaceXAI
- Priced at $2 per million input tokens and $6 per million output tokens
- Elon Musk: "It is an Opus-class model, but faster, more token-efficient, and lower cost"
- Musk added: "Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster"
- First major model release since SpaceXAI went public several weeks ago
How it compares on price:
- Anthropic Opus 4.7: $5 input / $25 output per million tokens
- OpenAI Sol: $5 input / $30 output per million tokens
- Grok 4.5: $2 input / $6 output per million tokens
Why it matters:
Grok 4.5 lands at the right moment - the same week OpenAI released GPT-5.6, and the AI market is actively debating whether premium model pricing is justified.
SpaceXAI is making a straightforward argument: comparable performance for a fraction of the cost. If the real-world results hold up, it gives businesses a concrete reason to at least test Grok alongside whatever they're currently paying more for.
👀Read more about Grok 4.5 updates!
🎙️ChatGPT Voice Now Listens While You Talk Without Interrupting
OpenAI launched GPT-Live, a new generation of voice models replacing Advanced Voice Mode, letting ChatGPT listen and speak at the same time for more natural back-and-forth conversation.
What's new:
- Rolling out globally on iOS, Android, and the ChatGPT web
- GPT-Live-1 becomes the default for Go, Plus, and Pro users
- GPT-Live-1 mini becomes the default for Free users
- API access coming soon
- Can interrupt, pause, or ask ChatGPT to slow down naturally
- Acknowledges what you're saying with phrases like "mhmm" or "got it"
- Waits when you pause to think instead of jumping in
- Filters out background noise like traffic or nearby conversations to focus on your voice
- Shows visual cards for weather, stocks, sports, and more while you talk
- Supports search, memory, images, and file uploads during voice conversations
- Delegates complex questions to GPT-5.5 in the background while keeping the conversation going
- Choose reasoning level: Instant for fast responses, Medium or High for deeper thinking
- Nine ChatGPT voices remastered for GPT-Live
- Over 150 million people already use ChatGPT Voice and Dictation weekly
Safety features:
- Safeguards act in real time while the model is speaking
- Age-appropriate behavior is trained directly into the model for teen users
- Parents can control teen access through Parental Controls
- Parents are notified in higher-risk situations involving signs of self-harm or suicidal intent
- Cannot impersonate real people's voices, uses predefined voice set only
- Crisis helpline support built into self-harm conversations
Limitations:
- Does not yet support voice with video or screen sharing
- Some languages may have a non-native accent or gaps in fluency
Why it matters:
The biggest shift here isn't that ChatGPT sounds more natural - it's that it can now handle complex tasks in the background while keeping the conversation going, which is the missing piece that made previous voice modes feel limited for anything beyond quick questions.
OpenAI's product lead has had 30 to 40-minute conversations using it on walks, which suggests this is less about a voice assistant and more about making AI genuinely usable hands-free for longer stretches of real work.
👀 Read more about ChatGPT’s new voice mode!
My Latest LinkedIn & X/Twitter Posts:
- 20 ChatGPT prompts for faster writing, planning & everyday work tasks (view post)
- How to use Microsoft Copilot across your workflows (view post)
- 15-step AI learning roadmap to go from beginner to expert (view post)
- Top 11 free AI tools from Google and how to use them (view post)
- 10 ChatGPT prompt templates to get better results faster (view post)
In partnership with Kit:
As a creator, your time should be spent doing what you love, not juggling a dozen tools just to run your business.
Kit gives you everything you need in one place:
✅ Build and grow your email list (I use Kit for my newsletter)
✅ Easily monetize with paid newsletters and digital products
✅ Automate your emails with triggers and custom workflows
✅ Track what’s working and optimize with powerful insights
It’s not just email. It’s your entire creator business, simplified.
Join a thriving community of successful creators.
Use Kit for Free — and Start Building Smarter
New AI Tools to Boost Your Productivity:
- WellSent: Fills out forms from your documents.
- Bgbye: Removes image backgrounds.
- Omago: Answers customer questions 24/7.
- InterMIND: Translates meetings and documents in real time.
- Avatar SDK: Creates 3D avatars from photos.
- Acousmos: Transcribes speech and saves searchable voice notes.
- LetterBucket: Creates and sends newsletters.
- Loova: Creates AI UGC ads from products.
- SophMate: Assists with WordPress and WooCommerce tasks.
- Grok 4.5: AI model for complex reasoning and coding tasks.
- MixTranslate: Compares translations from multiple AI models.
- VioEvo: Creates videos from text prompts.
- Video Transcription AI: Transcribes videos into text.
This Week's Top AI News Headlines:
- Google Makes AlphaEvolve, its AI System for Discovering and Optimizing Algorithms, Available to Developers Through Google Cloud (View Article)
- Anthropic Resets Hourly and Weekly Claude Usage Limits for All Users, Restoring Full Access After Rate Caps Were Reached (View Article)
- OpenAI Introduces New ChatGPT Workspace Features to Help Professionals Tackle Complex Projects, Long-Term Planning, and Collaborative Work (View Article)
- Anthropic Launches Reflect With Claude, an AI-Guided Journaling Tool Designed to Help Users Reflect, Organize Thoughts, and Support Personal Growth (View Article)
- Google Will Label AI-Generated Advertisements Across its Platforms, Giving Users Clearer Disclosure About Ads Created With Artificial Intelligence (View Article)
- CharacterAI, the Platform for Creating and Chatting With AI Characters, Launches Original AI-Powered Microdrama Series With Interactive Storytelling Features (View Article)
- Meta Launches Muse Spark 1.1, an AI Coding Assistant Designed to Compete With GitHub Copilot, Cursor, and Claude Code (View Article)
- Google Photos Launches Video Remix, an AI Feature That Transforms Personal Videos into Stylized Clips With Different Visual Effects and Themes (View Article)
- OpenAI Shuts Down Atlas Browser Project While Continuing to Expand its Long-Term Plans for AI-Powered Web Browsing (View Article)
- Microsoft Introduces ODE, a Specialized AI Model Designed to Generate Poetry With Greater Creativity, Literary Style, and Emotional Expression (View Article)
- Mistral AI Launches Robostral Navigate, a Vision-Language Model Designed to Help Robots Understand, Navigate, and Operate in Real-World Environments (View Article)
- Anthropic Adds Subtle Claude Prompts That Encourage Users to Rely More on AI Throughout Their Daily Workflows (View Article)
- OpenAI Says GPT-5.6 Remains the Preferred AI Model for Microsoft Copilot Despite Reports of a Growing Partnership Rift (View Article)
- Microsoft Reportedly Replaces Some OpenAI and Anthropic Models With Lower-Cost Alternatives to Reduce Enterprise AI Expenses (View Article)
- OpenAI Introduces New Coding Evaluation Framework to Better Measure Real-World AI Programming Skills by Reducing Misleading Benchmark Results (View Article)
Work With Me:
If you enjoy this newsletter, please forward it to your friends and colleagues.
Follow me on LinkedIn and X/Twitter to see my latest posts.
Have a wonderful week!
Andrew Bolis
|