Anthropic has released Claude Haiku 4.5, a new, smaller, and more affordable AI model designed for speed and efficiency. This model offers performance comparable to previous, larger versions, particularly in coding tasks, and is now available to all users via the Claude app, web, and API. For developers, Haiku 4.5 is priced at $1 per million input tokens and $5 per million output tokens. The model is also being integrated into GitHub Copilot for users on specific Pro, Pro+, Business, and Enterprise plans, with a public preview now available. This release positions Haiku 4.5 as a cost-effective option, making AI more accessible for a wider range of applications and free product versions. Meanwhile, the broader AI landscape sees new tools from Meta and OpenAI blurring the lines between news and advertising with realistic AI-generated videos, raising concerns about disinformation. In other developments, Nvidia's CEO highlighted the productivity gains from AI tools used by his company's engineers, while Ray Dalio is training an AI clone of himself to share investment insights. Microsoft faces a lawsuit alleging price manipulation through its investment in OpenAI, a claim Microsoft denies. Separately, the Newspaper Association of Korea is pushing for transparency in AI training data, and Djibouti is implementing AI training for its education inspectors.
Key Takeaways
- Anthropic has launched Claude Haiku 4.5, a faster and significantly cheaper AI model that matches the performance of older, larger models.
- Claude Haiku 4.5 is priced at $1/million input tokens and $5/million output tokens for API users.
- The new Haiku 4.5 model is available in a public preview for GitHub Copilot users on various paid plans.
- AI video tools from Meta and OpenAI are making it difficult to distinguish between news and advertising, raising concerns about disinformation.
- Nvidia's CEO noted that 100% of Nvidia's software engineers use an AI coding tool, highlighting productivity boosts.
- Ray Dalio is creating an AI clone of himself by training it on questions about financial markets.
- Microsoft is being sued for allegedly manipulating AI prices through its investment in OpenAI.
- The Newspaper Association of Korea is demanding transparency regarding the data used to train AI models.
- Djibouti is initiating AI training for its education inspectors as part of its digital transformation efforts.
- Consumers are increasingly using AI tools like ChatGPT and Gemini for shopping to find deals and save time, despite privacy concerns.
Anthropic releases Claude Haiku 4.5, a faster, cheaper AI model
Anthropic has launched Claude Haiku 4.5, a new artificial intelligence model that is smaller and less expensive than previous versions. This model is designed for users seeking quick and accurate responses. It performs comparably to larger models in coding tasks and is now the most affordable option for paid users. Anthropic's AI models are named Haiku for small, Sonnet for medium, and Opus for large, with Haiku 4.5 being the latest iteration. The company also announced Claude Sonnet 4.5 in September and Claude Opus 4.1 in August, with plans for another model update soon.
Anthropic offers cheaper AI model Haiku 4.5 for wider use
AI startup Anthropic has updated its smallest AI model, Haiku, to version 4.5. This new model is significantly cheaper, costing about one-third of its medium-sized Sonnet model and one-fifteenth of its top-tier Opus model. Anthropic states that Haiku 4.5 performs as well as or better than previous medium-sized models on various tests. This move aims to make AI systems more accessible and affordable for a broader range of users.
Anthropic's Claude Haiku 4.5 matches older AI performance at lower cost
Anthropic has released Claude Haiku 4.5, a small AI model that matches the performance of its previous top models but is much faster and cheaper. This new model is available to all users through the Claude app, web, and API. Haiku 4.5 is ideal for tasks like coding assistance due to its speed and lower cost, while larger models like Sonnet and Opus offer deeper analysis. For developers using the API, Haiku 4.5 is priced at $1 per million input tokens and $5 per million output tokens.
Claude Haiku 4.5 now in public preview for GitHub Copilot
Anthropic's latest AI model, Claude Haiku 4.5, is now available in a public preview for GitHub Copilot users. This model is designed for high performance and speed. It will be accessible to users on Copilot Pro, Pro+, Business, and Enterprise plans. Users can select Claude Haiku 4.5 within the Copilot Chat model picker in various GitHub and Visual Studio Code interfaces. Administrators may need to enable the policy for their organizations to access the new model.
Anthropic updates scaled-down Haiku AI model
Anthropic has released Claude Haiku 4.5, the newest version of its smallest AI model. This update offers performance similar to its Sonnet 4 model but at one-third the cost and more than twice the speed. Haiku 4.5 is available on all free Anthropic plans and is expected to be particularly useful for free AI product versions. The model's efficiency allows for parallel deployment or combination with more advanced models, enabling new production applications in areas like software development.
AI video tools blur lines between news and advertising
New artificial intelligence video tools are making it difficult to distinguish between authentic news reports and advertisements. Companies are using AI to create convincing videos featuring AI-generated anchors, victims, and even cloned influencers to promote services like legal aid. Tools from Meta and OpenAI allow for rapid creation of realistic AI videos, raising concerns about disinformation and the erosion of truth in advertising. Experts warn that these affordable tools can be misused to spread propaganda and scams, impacting the information ecosystem.
AI video tools blur lines between news and advertising
New artificial intelligence video tools are making it difficult to distinguish between authentic news reports and advertisements. Companies are using AI to create convincing videos featuring AI-generated anchors, victims, and even cloned influencers to promote services like legal aid. Tools from Meta and OpenAI allow for rapid creation of realistic AI videos, raising concerns about disinformation and the erosion of truth in advertising. Experts warn that these affordable tools can be misused to spread propaganda and scams, impacting the information ecosystem.
AI revolutionizes peptide design in biotech
Artificial intelligence is emerging as a powerful tool in biotechnology for discovering and designing peptides. Traditional methods for peptide discovery are time-consuming and resource-intensive due to the vast number of possible sequences. AI, using machine learning and deep learning, can analyze massive databases to identify patterns linking peptide sequences with biological activities. Companies like Nuritas and researchers at Duke University are developing AI platforms, such as the AI Magnifier and PepPrCLIP, to accelerate the identification and optimization of therapeutic peptides for various applications.
Semiconductor cooling tech can improve AI data center thermal management
The semiconductor industry's use of additive manufacturing (AM), or 3D printing, for advanced cooling solutions offers valuable lessons for AI infrastructure. Companies that build chip-making equipment have long used AM to create precise cooling components that manage temperature, vibration, and corrosion. This technology allows for complex geometries and integrated fluid flow control, enabling more efficient heat exchange in smaller footprints. Applying these AM techniques to data centers can help manage the high power density and heat generated by AI hardware, improving performance and chip quality.
Korean newspapers demand AI training data transparency
The Newspaper Association of Korea is calling for the disclosure of data used to train artificial intelligence models. They are urging South Korea to align with international standards, such as the European Union's AI Act, to ensure transparency. The association believes that knowing the training data is crucial for protecting content creators' rights and preventing copyright infringement and bias in AI outputs. They plan to work with government bodies and AI developers to establish clear regulations for data usage.
Nvidia CEO highlights AI startups and engineer tool use
Nvidia CEO Jensen Huang mentioned six AI startups that are shaping the future of workforces, combining humans and AI agents. He specifically noted that 100% of Nvidia's software engineers and chip designers use the AI coding tool Cursor, developed by Anysphere. Huang believes these AI agents significantly boost productivity and improve work quality. He envisions a future where companies like Nvidia employ millions of AI assistants alongside their human workforce to enhance output across all divisions.
Consumers use AI for savings and convenience
A new report shows that consumers are increasingly using artificial intelligence to save time and find better deals when shopping. Over a third of consumers prefer searching for products using AI compared to traditional search engines, with tools like ChatGPT and Gemini being the most popular. Many shoppers believe AI helps them find savings, discover better products, and reduce decision fatigue. While consumers value AI's benefits, they also express concerns about data privacy and the accuracy of AI recommendations.
Ray Dalio seeks market questions for AI clone training
Ray Dalio, founder of Bridgewater Associates, is inviting investors to ask him questions about financial markets. He plans to use these questions to train an AI clone of himself. Dalio believes this AI clone will help him share his investment insights and economic principles more broadly. This initiative aims to create a scalable way for Dalio's knowledge to reach a wider audience, allowing people to interact with his expertise on markets and investing.
Microsoft faces lawsuit over alleged AI price manipulation
Microsoft is being sued for allegedly manipulating artificial intelligence prices through its investment in OpenAI. The lawsuit claims that Microsoft used its influence to inflate OpenAI's ChatGPT prices significantly, potentially charging competitors 100 to 200 times more per token. It alleges a secret agreement allowed Microsoft to control compute supply and restrict OpenAI's output. Microsoft denies the claims, stating its partnership promotes competition and innovation. Experts suggest the lawsuit may raise policy concerns about AI market concentration but is unlikely to have a direct impact on Microsoft.
Djibouti launches AI training for education inspectors
Djibouti has launched a training program focused on artificial intelligence in education for its General Inspectorate of National Education (IGEN). This initiative, part of the 2025-2026 Continuing Education Plan, aims to enhance inspectors' skills in using AI ethically and effectively for supervision and decision-making. Led by Dr. Houssein Ahmed Assowe, the program includes theoretical modules and workshops covering digital ethics, data management, and adaptive learning. Scheduled from October 2025 to January 2026, this training is a key step in Djibouti's digital transformation of its education system.
Sources
- Anthropic launches Claude Haiku 4.5, a smaller, cheaper AI model
- US tech startup Anthropic unveils cheaper model to widen AI's appeal
- Anthropic’s Claude Haiku 4.5 matches May’s frontier model at fraction of cost
- Anthropic’s Claude Haiku 4.5 is in public preview for GitHub Copilot
- Anthropic launches new version of scaled-down "Haiku" model
- Is that a newscast or a sales pitch? New AI videos make it tough to tell
- Is that a newscast or a sales pitch? New AI videos make it tough to tell
- The advent of AI for peptide design: An emerging field
- What Semiconductor Cooling Can Teach AI Infrastructure About Precision Thermal Management
- Newspaper Association Demands AI Training Data Disclosure
- Jensen Huang name-checks 6 AI companies — and says 100% of Nvidia engineers use one of them
- Consumers embracing AI to save time, find deals
- Ray Dalio wants X users to ask him questions about markets to train his AI clone
- Microsoft accused of trying to manipulate AI pricing via its OpenAI investment
- Djibouti Embarks on AI Education Revolution with Inspectorate Training Program