Why ChatGPT‑5 Feels Smarter in Some Ways — and Confusing in Others
Understanding the updates, capabilities, and limitations of OpenAI’s latest model
OpenAI’s latest language model, ChatGPT‑5, brings updates in context handling, multimodal input, and task-specific modes.
These improvements promise more powerful AI interactions, but user feedback shows both strengths and quirks.
What’s New
Extended Context Handling: ChatGPT‑5 can track longer conversations without losing prior information.
Multimodal Capabilities: The model can understand and generate text, images, and code.
Mode-Based Reasoning: Specialized modes help tackle tasks like research, problem-solving, or creative writing.
Users report these features improving workflows in writing, coding, and research.
Where It Struggles
Multi-step logic errors and complex problem-solving mistakes.
Responses that are internally consistent but incorrect against external sources.
Inconsistent answers when switching between tasks or topics.
Key Takeaways
ChatGPT‑5 introduces exciting key updates, but reasoning errors and variable performance remain. OpenAI is responding with improvements:
Warmer, more approachable responses.
Restoring GPT-4o access for paid users.
Improved model switching for more reliable performance.
📚 Bibliography & Sources
News & Media Coverage
The Washington Post (Aug 17, 2025). The new ChatGPT has some AI fans rethinking when to expect 'superintelligence'.
https://www.washingtonpost.com/technology/2025/08/17/openai-gpt5-chatgpt-superintelligenceIndiaTimes (Aug 2025). Is ChatGPT 5 factually inaccurate or biased? Paid version 'thinks' Joe Biden, not Donald Trump, is the US President; users outrage
https://www.indiatimes.com/technology/is-chatgpt-5-factually-inaccurate-or-biased-paid-version-thinks-joe-biden-not-donald-trump-is-us-president-users-outrage-666984.htmlEl País (Aug 17, 2025). Modo 'beta' global: el experimento masivo de la IA.
https://elpais.com/tecnologia/2025-08-17/modo-beta-global-el-experimento-masivo-de-la-ia.htmlEl País (Cinco Días) (Aug 11, 2025). GPT-5 de OpenAI: estas son sus grandes novedades…
https://cincodias.elpais.com/smartlife/lifestyle/2025-08-11/gpt-5-openai-grandes-novedades.htmlTom’s Guide (Aug 2025). What is ChatGPT-5? – new features, how to use it, plans, pricing and more.
https://www.tomsguide.com/ai/what-is-chat-gpt-5Windows Central (Aug 2025). Google took the stage for OpenAI's GPT-5 launch — but Microsoft services quietly got the biggest boost
https://www.windowscentral.com/artificial-intelligence/openai-chatgpt/google-took-the-stage-for-openais-gpt-5-launch-but-microsoft-services-quietly-got-the-biggest-boostThe Guardian (Aug 2025). Nvidia, AMD, Trump, and OpenAI’s GPT-5: A new wave of AI hype.
https://www.theguardian.com/global/2025/aug/11/nvidia-amd-trump-openai-gpt-5-aol-dial-upThe Verge (Aug 2025). GPT-5 failed the hype test.
https://www.theverge.com/openai/759755/gpt-5-failed-the-hype-test-sam-altman-openaiTom’s Guide (Aug 2025). ChatGPT-5 is here – 7 biggest upgrades you need to know.
https://www.tomsguide.com/ai/chatgpt-5-is-here-openais-most-powerful-ai-model-just-changed-everythingIndiaTimes (Aug 2025). OpenAI launches ChatGPT-5 and the new version might just replace your therapist, tutor, and best friend.
https://www.indiatimes.com/trending/openai-launches-chatgpt-5-and-the-new-version-might-just-replace-your-therapist-tutor-and-best-friend-heres-why-666206.htmlStatesman (Aug 2025). OpenAI releases new AI model ChatGPT-5. Here's what's new, when it's available
https://www.statesman.com/business/technology/article/chatgpt-5-released-update-openai-sam-altman-20807199.phpIndian Express (Aug 2025). Sam Altman says OpenAI ‘screwed up’ GPT-5 rollout: Here are the changes made so far. https://indianexpress.com/article/technology/artificial-intelligence/sam-altman-openai-screwed-up-gpt-5-rollout-here-are-changes-10194555/
Academic & Technical Research
Garousi, V. (Apr 2025). Why you shouldn't fully trust ChatGPT: A synthesis of this AI tool's error rates across disciplines and the software engineering lifecycle
https://arxiv.org/abs/2504.18858Groza, A. (Oct 2023). Measuring reasoning capabilities of ChatGPT.
https://arxiv.org/abs/2310.05993Mündler, N., He, J., Jenko, S., & Vechev, M. (May 2023). Self-contradictory Hallucinations of Large Language Models.
https://arxiv.org/abs/2305.15852Chen, L., Zaharia, M., & Zou, J. (Jul 2023). How is ChatGPT's behavior changing over time?
https://arxiv.org/abs/2307.09009
Background & History
Wikipedia (2025). GPT-5.
https://en.wikipedia.org/wiki/GPT-5Cursor IDE Blog (2025). What’s New in ChatGPT-5: A Comprehensive Guide.
https://www.cursor-ide.com/blog/whats-new-chatgpt-5Medium (Pimrada Thavorn) (2025). Launch of ChatGPT-5: Breaking Boundaries in AI Power.
https://medium.com/@GizmoJournalDetails/launch-of-chatgpt-5-breaking-boundaries-in-ai-power-cd67a88e2421Wikipedia (2025). 2025 in artificial intelligence.
https://en.wikipedia.org/wiki/2025_in_artificial_intelligenceWikipedia (2025). ChatGPT Deep Research.
https://en.wikipedia.org/wiki/ChatGPT_Deep_Research
I like your newsletter, but you use a lot of AI-generated images. I can't explain it, but AI images have become such a turnoff that it makes me not want to read the content. Just one person's opinion.