Tech News Hub
OpenAI Enhances ChatGPT with Powerful Image-Generation Capabilities
OpenAI has introduced a significant upgrade to ChatGPT’s image-generation features, marking the first major improvement in over a year. This new capability is powered by the company’s GPT-4o model, expanding ChatGPT’s abilities beyond text-based responses to include native image creation and editing.
Key Features of GPT-4o's Image Generation
- Enhanced Image Creation: GPT-4o generates highly detailed and accurate images, improving upon the previous DALL-E 3 model. While GPT-4o may take slightly longer to produce results, OpenAI claims this ensures better precision and realism.
- Advanced Editing Tools: Users can modify existing images directly within ChatGPT, including adjusting backgrounds, transforming elements, or refining intricate details.
- Support for Realistic Human Imagery: GPT-4o’s capabilities extend to editing images featuring people, allowing for enhanced photo adjustments while maintaining quality.
Availability
The upgraded image-generation feature is now available to Pro subscribers as part of OpenAI’s $200-a-month plan. OpenAI plans to expand access to Plus and free users soon, as well as developers using its API service.
Training Data and Copyright Measures
To train GPT-4o, OpenAI used publicly available data alongside proprietary content from partnerships with platforms like Shutterstock. OpenAI’s COO, Brad Lightcap, emphasized that the company enforces strict policies to prevent generating content that mimics living artists' work.
For added transparency, OpenAI offers an opt-out form for creators who wish to exclude their content from future training datasets. The company also respects requests from website owners to prevent its web-scraping bots from gathering data.
Competitive Landscape
This update follows Google’s recent launch of image-generation capabilities for its Gemini 2.0 Flash model, which faced criticism for lacking safeguards that allowed the creation of copyrighted content. OpenAI’s stricter policies aim to prevent similar issues.
Conclusion
With GPT-4o’s improved image-generation capabilities, ChatGPT takes a major step forward in creative content production. By combining detailed image creation with advanced editing tools, OpenAI aims to deliver a versatile solution that balances innovation with responsible content practices.
Related Stories:
Top Stories:
OpenAI Introduces O1-Pro: Its Most Powerful and Costly AI Model Yet
OpenAI has released O1-Pro, a more advanced version of its O1 "reasoning" AI model, as accessible now via its developer API. The new model is aimed at making more accurate and stable responses with the use of greater computational power.
How Does O1-Pro Differ?
O1-Pro builds on the capabilities of the original O1 model but is designed to handle complex problems with improved reasoning skills. OpenAI claims this version offers “consistently better responses,” thanks to its ability to process information more thoroughly.
Limited Access with Premium Pricing
For now, O1-Pro is available only to developers who have spent at least $5 on OpenAI’s API services. However, its pricing is notably higher than previous models:
- $150 per million tokens (roughly 750,000 words) for input
- $600 per million tokens for output
Twice that of GPT-4.5 for input and ten times greater than the basic O1 model.
Why Such a Pricey Tag?
OpenAI is billing O1-Pro as an upmarket offering for developers that require improved accuracy and dependability for sophisticated questions. OpenAI describes O1-Pro as deploying more computing muscle to provide superior outcomes, especially for difficult problem-solving situations.
Mixed Reviews from Early Users
Early users have been delivering mixed reviews. Since December, ChatGPT Pro subscribers have had access to O1-Pro, and some users reported the model having a hard time completing tasks such as Sudoku puzzles and deciphering optical illusion jokes.
Internal tests from OpenAI suggest that while O1-Pro slightly outperforms the standard O1 in coding and math tasks, its main strength lies in delivering more consistent responses for complex requests.
Conclusion
O1-Pro’s introduction reflects OpenAI’s focus on providing advanced AI capabilities for developers — but at a significantly higher cost. Whether the model’s improved consistency and reasoning justify its steep price remains to be seen.
Related Stories:
Top Stories:
OpenAI to Begin Beta Testing ChatGPT Connectors for Google Drive and Slack
OpenAI is releasing a new feature known as ChatGPT Connectors that will enable business users to link their Google Drive and Slack accounts with ChatGPT. The integration will enable ChatGPT to respond to questions based on company files, presentations, spreadsheets, and Slack conversations from a single platform.
The feature will initially be rolled out in beta to a select group of ChatGPT Team members, with a future expansion to platforms such as Microsoft SharePoint and Box.
Driven by a customized version of GPT-4o, ChatGPT Connectors can fine-tune its response based on internal company information. In order to maintain confidentiality, OpenAI has stressed that Slack and Google Drive permissions will be stringently adhered to — employees have access to content only they already have permission to see.
There are a few caveats, though. ChatGPT Connector won't scan Google Drive file pictures, and it won't be able to see Slack DMs, group messages, or bot threads. Admins will be able to dictate which Slack channels and Google Drive files are included too.
To become a beta program participant, interested businesses will have to submit 100 test documents, spreadsheets, presentations, or Slack conversations. OpenAI has promised users that data synced will not be used directly for training ChatGPT but could be utilized to generate synthetic data for model updates.
Through ChatGPT Connectors, OpenAI intends to make it easier for businesses to access information while staying committed to robust data security practices.
Related Stories:
Top Stories:
OpenAI Unveils New Tools to Help Businesses Build Smarter AI Agents
OpenAI has introduced a set of new tools designed to help developers and businesses create more advanced AI agents. These tools are part of OpenAI’s newly launched Responses API, which aims to improve how organizations build automated systems that can perform complex tasks independently.
What’s New with the Responses API?
The Responses API is designed to replace OpenAI’s previous Assistants API, which the company plans to retire by mid-2026. This updated API allows developers to build AI agents capable of performing web searches, navigating websites, and scanning company files for information.
This API is based on OpenAI’s Operator product, offering developers more flexibility to create customized AI solutions for business needs.
Enhanced AI Models for Better Results
The Responses API introduces two new AI models that focus on improved accuracy and performance:
- GPT-4o search – Achieves a 90% accuracy score on OpenAI’s SimpleQA benchmark.
- GPT-4o mini search – Scores 88% on the same benchmark.
Both models are designed to browse the web for information and provide sources in their responses. These models outperform OpenAI’s recently introduced GPT-4.5 model, which scored 63%.
File Search and Automated Actions
The Responses API also includes other features to boost productivity:
- File Search Utility – A utility that rapidly searches company databases for pertinent information. OpenAI guarantees that information from this utility will not be used to train its models.
- Computer-Using Agent (CUA) – A simulation model that replicates mouse and keyboard movements to automate activities like data entry and app workflows. The CUA model is also offered in research preview and can be executed locally for companies looking for more control.
Overcoming Challenges in AI Agents
Though the Responses API is potent in tools, OpenAI admits that AI agents have yet to overcome all limitations. AI-driven search features are still liable to make occasional mistakes, with GPT-4o search continuing to answer about 10% of fact-based questions incorrectly. Moreover, citations in responses generated by AI may not always be trustworthy.
OpenAI has termed such tools as beta versions with future updates promising improvements.
Open-Source Tools for Developers
To support developers, OpenAI is also launching the Agents SDK, an open-source toolkit that helps with:
- Integrating safeguards for AI agents
- Monitoring agent activities for optimization
- Debugging tools to improve agent performance
This toolkit builds on OpenAI’s Swarm framework, which was introduced last year for managing multiple AI agents.
The Future of AI Agents
OpenAI API product chief Olivier Godement recognized the difficulty of scaling AI agents efficiently and maintaining consistent performance. He stressed that OpenAI is aiming beyond flashy demos to provide useful tools that companies can depend on.
With OpenAI CEO Sam Altman forecasting that 2025 will be the year of the ascent of AI agents into the labor force, these new technologies mirror the company's concentration on converting promise into actualized impact.
Related Stories:
Top Stories:
Microsoft Expands AI Ambitions: Developing New Models to Rival OpenAI
Microsoft is stepping up its AI game, developing its powerful models to compete with OpenAI — a company it has heavily invested in.
Reports suggest Microsoft has created its own AI "reasoning" models similar to OpenAI’s o1 and o3-mini. Tensions between the two firms reportedly rose after OpenAI declined to share technical details about its o1 model.
In addition, Microsoft has introduced a family of models known as MAI, which are said to perform on par with OpenAI’s technology. There’s speculation that Microsoft may offer these models through an API later this year.
To diversify its AI strategy, Microsoft is also testing models from xAI, Meta, Anthropic, and DeepSeek as potential alternatives for powering its Copilot assistant.
Despite investing around $14 billion in OpenAI, Microsoft appears to be preparing for greater independence in the AI space. Notably, the company recently hired Mustafa Suleyman, co-founder of DeepMind and Inflection, to spearhead its AI initiatives.
Related Stories:
Top Stories:
ChatGPT for macOS Can Now Directly Edit Code Inside Developer Tools
OpenAI has brought a significant update for macOS users—ChatGPT can now edit code directly inside developer tools such as Xcode, VS Code, and JetBrains. This avoids the inconvenience of copying and pasting, and coding becomes smoother and more efficient.
Auto-Apply Mode for Quicker Edits
To accelerate the workflow, ChatGPT now has an auto-apply mode, which enables changes to be applied immediately without the need for approval. The feature is supported by ChatGPT Plus, Pro, and Team users, with a wider release to Enterprise, Edu, and free users later this week.
Windows Support Coming Soon
For the time being, this feature is available only on macOS, but OpenAI has assured that Windows users will receive access in the near future, providing cross-platform support for developers.
Competing with GitHub Copilot and Cursor
This update expands on OpenAI’s "work with apps" feature, which launched in beta in November 2024, enabling ChatGPT to interact with code directly inside development environments. With this new editing capability, ChatGPT is becoming a strong rival to AI coding assistants like GitHub Copilot and Cursor.
AI Coding: The Benefits and Challenges
Code-generation tools powered by AI are also becoming increasingly popular, with the most recent GitHub survey showing that the majority of developers now make use of AI in their processes. Yet, difficulties persist—a report by Harness revealed that code created by AI usually needs more debugging, while Google identified that AI accelerates code reviews but has the potential to affect software stability at times.
The Future of AI-Powered Development
OpenAI is also said to be developing a special AI tool for software developers that could further change how developers write and optimize code. As AI continues to improve at a rapid pace, ChatGPT's recent additions herald the trend toward more intelligent and automated software development.
Related Stories:
Top Stories:
OpenAI Rolls Out GPT-4.5 to More ChatGPT Users – What You Need to Know
OpenAI is rolling out broader access to its newest AI model, GPT-4.5, offering it to a larger group of users on the ChatGPT Plus plan. First available for ChatGPT Pro subscribers, the expansion is now rolling out to more people. OpenAI posted on X (formerly Twitter) that it would take between one to three days, and rate limits might change as usage builds up.
What's New in GPT-4.5?
GPT-4.5 is the largest and most sophisticated AI model of OpenAI to date, developed with greater computing power and a larger dataset than its predecessors. It includes enhancements in knowledge depth, emotional intelligence, and lower hallucination rates, which make it more trusted in producing accurate answers.
However, despite these advancements, GPT-4.5 is not necessarily OpenAI’s strongest model across the board. Competing AI models from DeepSeek, Anthropic, and even OpenAI have performed better in specific reasoning tasks.
The High Cost of GPT-4.5
Operating GPT-4.5 is expensive, leading OpenAI to reconsider whether it will remain available through its API. To manage costs, the company has set premium pricing for its use:
- $75 per million tokens (about 750,000 words) for input
- $150 per million tokens for output
This makes it significantly more expensive than OpenAI’s GPT-4o, which remains the company’s most cost-efficient model.
A Highly Persuasive AI Model
One of the more interesting aspects of GPT-4.5 is how it can influence others. The internal tests from OpenAI found that the model was very skilled at persuading other AIs, even compelling them to leak sensitive information or transfer virtual currencies. This illustrates its potential within applications that necessitate sophisticated thinking and strategic interaction.
What's Next?
As OpenAI further perfects its AI models, the long-term future of GPT-4.5 is unsure because of its very high operating costs. With wider availability now open to ChatGPT Plus users, more individuals can try out its capabilities and compare them to earlier models.
Related Stories:
Top Stories:
