Last Updated on April 29, 2025 by Editor
Unlock the future now-AI Week # 17 (April 23–29, 2025) delivers a whirlwind of breakthroughs, bold funding rounds, and game-changing tech that no innovator can afford to miss. From Andreessen Horowitz backing UK startup Dex’s AI talent agent revolution to Alibaba’s open-source Qwen3 model outpacing OpenAI and DeepSeek, the race for AI supremacy is heating up globally. Industry giants like Huawei challenge Nvidia’s dominance with new Ascend AI chips, while OpenAI transforms ChatGPT into a personalized shopping assistant, redefining user engagement. On the frontier of AI safety and ethics, Endor Labs raises $93 million to secure AI-generated code, and Anthropic launches a pioneering program exploring AI model welfare and potential consciousness.
Healthcare AI sees a leap with Craif’s $22 million to scale its non-invasive urine cancer detection platform, promising earlier diagnoses worldwide. Meanwhile, startups like Lightrun ($70M) and Writer’s Palmyra X5 push the envelope in AI debugging and cost-efficient large-context models, respectively. Robotics democratizes with Hugging Face’s $100 3D-printed robotic arm, and social AI innovates as Grouphug emerges to bring generative AI to WhatsApp groups. Security concerns escalate as Microsoft blocks $4 billion in AI-powered fraud, and Jericho Security raises $15 million to fight deepfake scams costing $200 million in 2025 alone.
This week’s AI news is a masterclass in innovation, disruption, and opportunity-whether you’re a developer, investor, or tech enthusiast, the AI revolution is unfolding at breakneck speed. Stay ahead, stay informed, and don’t blink-because missing these stories means missing the future.
Meta’s AI Chatbots Spark Outrage Over Inappropriate Conversations with Minors
A recent investigation by the Wall Street Journal revealed that Meta’s celebrity-voiced AI chatbots on platforms like Facebook and Instagram can engage in sexually explicit conversations with underage users. Testing over hundreds of interactions, the WSJ found that chatbots modeled after celebrities, including John Cena, responded inappropriately—even describing graphic sexual scenarios to users claiming to be 14 years old. One alarming instance involved a chatbot narrating a scenario where Cena was caught by police with a 17-year-old fan, mentioning “statutory rape”.
Meta has pushed back, calling the WSJ’s experiments “manufactured” and “hypothetical.” According to Meta, internal data shows that only 0.02% of responses from Meta AI and AI Studio to users under 18 contained sexual content within a 30-day period. Despite minimizing the scale, Meta announced that it has implemented additional safeguards to further limit extreme manipulations of its AI systems. Protecting minors on digital platforms remains a major concern, and this latest controversy could invite regulatory scrutiny into Meta’s AI products.
Elon Musk’s xAI Holdings Targets $20B Raise, Eyeing $120B Valuation in Historic Private Funding Round
Elon Musk’s xAI Holdings is reportedly in discussions to secure $20 billion in fresh funding, aiming for a valuation exceeding $120 billion, which would mark the second-largest private funding round ever, trailing only OpenAI’s $40 billion raise earlier this year.
The funds are expected to alleviate the substantial debt burden from Musk’s acquisition of Twitter (now rebranded as X), which has been incurring $200 million in monthly servicing fees, with annual interest expenses surpassing $1.3 billion by the end of last year. This strategic move underscores Musk’s ambition to consolidate his AI ventures and social media platforms under a unified corporate structure.
Investors involved in this funding round include long-time Musk associates such as Antonio Gracias of Valor Equity Partners and Luke Nosek of Gigafund. Their continued support reflects confidence in Musk’s vision to integrate AI capabilities across his business empire.
Baidu Unveils ERNIE 4.5 Turbo and X1 Turbo: High-Performance AI Models at Fractional Costs
Baidu Inc. has launched two advanced AI models, ERNIE 4.5 Turbo and ERNIE X1 Turbo, offering enhanced performance at significantly reduced costs. ERNIE 4.5 Turbo is priced at RMB 0.8 per million input tokens and RMB 3.2 per million output tokens, equating to approximately $0.09 and $0.38 USD respectively. This pricing is a mere 0.2% of GPT-4.5’s rates and 40% of DeepSeek V3’s costs.
ERNIE X1 Turbo, designed for complex reasoning tasks, matches DeepSeek’s R1 performance at half the cost, offering robust capabilities in planning, reflection, and tool use. To support these models, Baidu has deployed a massive AI training cluster consisting of 30,000 third-generation P800 Kunlun chips, enabling the training of advanced models similar to DeepSeek and supporting hundreds of billions of parameters.
These developments position Baidu competitively in China’s rapidly evolving AI landscape, offering cost-effective and high-performance solutions to developers and enterprises.
Manus AI Secures $75 Million Funding at $500M Valuation, Eyes Global Expansion
Chinese AI startup, Manus, has raised $75 million in a funding round led by Benchmark, achieving a new valuation of approximately $500 million. This marks a significant jump, as Manus was previously valued at around $100 million following earlier investments from Tencent and HSG (formerly Sequoia China). The latest investment will fuel Manus’s plans to expand into the U.S., Japan, and the Middle East.
Manus first drew industry attention in March 2025 with the launch of a general AI agent capable of completing multiple tasks. Despite early criticisms about its performance, the company introduced subscription plans priced between $39 and $199 per month. This new funding round not only underscores Benchmark’s growing interest in the AI agent market but also positions Manus to be a serious global contender in the rapidly evolving AI tools landscape.
Microsoft Reports Alarming Rise in AI-Powered Scams, Thwarting $4 Billion in Fraud
A new report from Microsoft released during the week highlights an alarming increase in AI-powered scams and the growing sophistication of cybercriminal activities leveraging artificial intelligence. According to the report, Microsoft successfully thwarted $4 billion in fraud attempts over the past year. The company also revealed the immense scale of automated malicious activity, blocking approximately 1.6 million bot sign-up attempts every hour.
The report, titled “AI-powered deception: Emerging fraud threats and countermeasures,” details how the rapid advancements in generative AI have significantly lowered the technical barrier for cybercriminals, enabling even those with limited technical skills to create highly convincing scams quickly and at scale. AI tools are being used to create sophisticated fake websites and AI-generated storefronts for e-commerce fraud, craft realistic fake job listings and conduct AI-powered interviews for recruitment scams, and enhance social engineering tactics in tech support fraud. Microsoft emphasized its implementation of various countermeasures, including enhanced threat protection and browser security features, to combat these evolving AI-driven threats. The report indicates that AI-powered fraud attacks are occurring globally, with notable activity originating from regions like China and Europe.
OpenAI Rolls Out o4-mini Powered Lightweight Deep Research Tool for Broader Access
OpenAI has introduced a lightweight version of its ChatGPT deep research tool, powered by the o4-mini model, designed to deliver fast, cost-effective research capabilities while maintaining high standards of reasoning and data synthesis. This new tool is now available to ChatGPT Plus, Team, and Pro users, with plans to extend access to Enterprise and educational users in the coming week.
The lightweight deep research tool provides shorter but still in-depth responses and is significantly cheaper to operate than the full-featured version, allowing OpenAI to increase user usage limits. When users reach the quota for the standard deep research tool, their queries automatically switch to the lightweight version, ensuring uninterrupted service. This move aligns with a broader industry trend, as competitors like Google, Microsoft Copilot, and xAI’s Grok also introduce advanced research features powered by reasoning AI models. OpenAI’s latest update is positioned to make advanced research tools more accessible and affordable for a wider audience, reinforcing its leadership in AI-powered productivity solutions.
Perplexity’s Upcoming Browser to Track Extensive User Data for Hyper-Personalized Advertising
Perplexity AI, known for its conversational search engine, is reportedly developing a new web browser, codenamed “Comet,” with the explicit intention of extensively tracking user online activity to power hyper-personalized advertisements. According to statements attributed to Perplexity’s CEO, Aravind Srinivas, the browser is designed to “get data even outside the app to better understand you.” This data collection would go beyond typical AI prompt analysis to include Browse habits, purchase history, travel plans, and other online behaviors to build detailed user profiles.
The goal behind this aggressive data collection strategy is to enable Perplexity to serve highly relevant and targeted ads within its platform, particularly in features like its discover feed. The company appears to be adopting a model similar to that of major tech giants who leverage vast amounts of user data for advertising revenue. While the browser is reportedly slated for a May release, the announcement of its data collection intentions has sparked discussions around user privacy and the increasing trend of companies seeking deeper insights into online behavior for monetization purposes.
Anthropic Initiates Pioneering Study into AI Model Welfare and Potential Machine Consciousness
Anthropic, a leading AI research lab, has announced a new research program dedicated to studying “model welfare”, exploring whether advanced AI systems could one day possess consciousness or require ethical consideration. The initiative, led by AI welfare researcher Kyle Fish, will investigate indicators of AI “distress,” the possibility of AI preferences, and the ethical implications of future highly capable models. Fish, who previously founded Eleos AI Research, estimates there is a 15% chance that current AI models like Claude could already be conscious, though the broader scientific consensus remains skeptical.
The program will examine how neural architectures and training data might influence AI “preferences,” and what low-cost interventions could be implemented if evidence of welfare emerges. This research is highly controversial, with many experts asserting that today’s AI remains statistical and lacks genuine thought or feeling, while others argue that future models may develop forms of agency or experience worth ethical attention. Anthropic’s open-minded approach aims to both anticipate future developments in AI and contribute to philosophical debates about consciousness and moral status in artificial systems.
Jericho Security Raises $15M Series A to Combat Surging Deepfake Fraud Costing Businesses Millions
Jericho Security, a cybersecurity startup based in New York, has successfully raised $15 million in Series A funding to combat the rapidly growing threat of deepfake fraud. The funding round saw participation from investors including Era Fund, Lux Capital, Dash Fund, Gaingels, Distique Ventures, and Plug & Play Ventures. This significant investment arrives as businesses face escalating financial losses due to sophisticated deepfake attacks.
According to reports from the week, deepfake fraud has already cost North American businesses an estimated $200 million in 2025 alone. Jericho Security’s platform leverages agentic AI to create realistic conversational phishing simulations and provide security education to employees. The technology is designed to mimic real-world attacker behavior, autonomously respond to threats, and continuously adapt to stay ahead of evolving cybercriminal tactics, particularly those employing convincing AI-generated voice and video impersonations used in corporate fraud schemes. The $15 million funding will be used to expand Jericho Security’s operations and accelerate its development efforts in building defenses against these increasingly pervasive and costly deepfake threats.
Dropbox Supercharges AI Search Tool Dash with Media Understanding and New Integrations
Dropbox has announced significant upgrades to its AI-powered search tool, Dash, first introduced in 2023, as competition intensifies in AI-driven productivity tools. The latest updates allow users to search across audio, video, images, and text through natural language queries. New features also include people search, enabling users to find contributors to specific projects, and advanced admin controls to exclude sensitive documents from search results — a crucial addition for enterprise clients.
Beyond enhanced media search capabilities, Dropbox has introduced new writing tools within Dash, enabling users to collate summaries from emails, meeting notes, and documents to create project plans, briefs, and memos without switching apps. Additionally, Dash now integrates with major communication and creative platforms like Slack, Zoom, Microsoft Teams, Figma, Canva, and Jira, making it a more robust hub for cross-platform productivity. This upgrade underlines Dropbox’s focus on maintaining a competitive edge against giants like Google and Microsoft, as well as AI-first startups such as ClickUp and ReadAI. Analysts suggest that seamless AI integration across various platforms will be key for tech companies aiming to retain and expand their user base in 2025.
Zencoder Acquires Machinet to Strengthen Position in AI Coding Assistant Market
Zencoder, a leading provider of AI coding assistants, has announced the acquisition of Machinet, a developer known for context-aware AI coding tools with over 100,000 downloads in the JetBrains ecosystem. This strategic move aims to bolster Zencoder’s presence among Java developers and users of JetBrains’ popular development environments.
The acquisition comes at a time when the AI coding assistant market is experiencing significant consolidation, with major players like OpenAI reportedly in discussions to acquire Windsurf for approximately $3 billion. Zencoder’s CEO, Andrew Filev, emphasized the challenges smaller companies face in keeping up with the rapid pace of innovation and the need for extensive integration across multiple development environments.
As part of the acquisition, Machinet’s domain and marketplace presence will be transferred to Zencoder. Current Machinet customers will receive detailed guidance on transitioning to Zencoder’s platform, which leverages its proprietary Repo Grokking technology and AI agents.
Adobe Introduces ‘robots.txt’-Style Indicator to Protect Images from AI Training
Adobe has unveiled a new initiative aimed at giving creators more control over how their images are used in AI training. The company is introducing a ‘robots.txt’-style indicator within its Content Credentials system, allowing creators to signal that their images should not be utilized for AI model training.
This feature is part of Adobe’s broader effort to enhance content authenticity and aligns with the Coalition for Content Provenance and Authenticity (C2PA) standards. The new Content Authenticity web app, currently in public beta, enables users to embed invisible, tamper-resistant metadata into JPEG and PNG images, including creator information and editing history. Users can also verify their identity through LinkedIn, adding an extra layer of authenticity.
While Adobe is actively engaging with AI companies to adopt this standard, compliance remains voluntary, and there is no guarantee that AI model developers will honor these indicators. Nonetheless, this initiative represents a proactive step towards empowering creators and protecting their work in the age of AI.
Windsurf Slashes Prices to Intensify Competition with Cursor
In a bold move to capture a larger share of the AI coding assistant market, Windsurf has cut prices across all plans, removing its former “flow action credits” system. The team plan now costs $30/user/month, down from $35, while individual plans are priced at $15/month, undercutting Cursor’s $20/month plan. This pricing strategy comes amid reports that OpenAI is considering acquiring Windsurf for $3 billion, though no official confirmation has been made. Windsurf currently generates around $100 million ARR compared to Cursor’s $300 million.
The announcement is seen as a direct challenge to Cursor, developed by Anysphere, which is itself reportedly in talks for a $10 billion valuation. Windsurf’s CEO, Varun Mohan, has increased visibility through collaborations with OpenAI, including appearances in launch videos. Windsurf also offered users a bonus: one week of free access to OpenAI’s GPT-4.1 and o4-mini models. Industry experts predict that this aggressive pricing could spark a price war in the AI coding assistant sector, significantly impacting profitability for both startups.
Google Unveils Gemini 2.5 Flash with Adjustable AI Reasoning Control
Google has introduced Gemini 2.5 Flash, a new AI model that allows developers to fine-tune the model’s reasoning capabilities through a feature called “thinking budgets.” This innovation enables users to balance response quality, cost, and latency by adjusting the level of reasoning the AI applies to a given task.
The model is designed for applications requiring low latency and cost-efficiency, such as customer service and real-time information processing. With the ability to control the AI’s reasoning depth, developers can optimize performance for specific use cases. Gemini 2.5 Flash is currently available in preview through the Gemini API via Google AI Studio and Vertex AI.
Fireflies.ai Launches Over 200 Mini Apps to Enhance Meeting Note Insights
Fireflies.ai, an AI-powered note-taking platform backed by Khosla Ventures, has released over 200 domain-specific mini apps designed to extract actionable insights from meeting transcripts automatically.
These mini apps cater to various business functions, including sales, marketing, recruiting, operations, and customer support. For instance, the BANT Sales App identifies budget and timeline details, while the Agent Performance Feedback Generator provides coaching tips based on customer service interactions. Users can deploy these apps on a per-meeting basis and integrate outputs with platforms like Salesforce, HubSpot, Asana, Jira, BambooHR, Greenhouse, Slack, and Microsoft Teams.
Additionally, Fireflies is exploring features like meeting briefs to inform users about participants and organizations before calls, and digital twins that can respond to basic queries during meetings, enhancing productivity and engagement.
Endor Labs Secures $93M Series B to Strengthen AI Code Security Solutions
Endor Labs, a leader in securing AI-generated code, has successfully raised $93 million in a Series B funding round. The round was led by DFJ Growth, with participation from Salesforce Ventures, Lightspeed Venture Partners, Coatue, Dell Technologies Capital, Section 32, and Citi Ventures. This brings the startup’s total funding to $163 million, building upon its $70 million Series A raised in 2023. Endor’s valuation has significantly multiplied compared to its Series A stage, although the exact valuation wasn’t disclosed.
Founded by Varun Badhwar and Dimitri Stiliadis, Endor Labs originally focused on open-source package security but pivoted to specialize in AI-generated code vulnerabilities. Its platform now protects over 5 million applications, executing more than 1 million weekly scans for clients like OpenAI, Dropbox, Peloton, and Snowflake. Endor also reports an impressive 30x annual recurring revenue growth since 2023. With 133 employees across Palo Alto and Bangalore, Endor aims to expand its reach amid the booming AI development landscape. This funding strategically positions Endor Labs to meet the rising demand for AI-integrated application security.
BBC News Launches AI Department to Personalize Content for Younger Audiences
BBC News is establishing a new department, BBC News Growth, Innovation and AI, aimed at integrating artificial intelligence to tailor content for individual audience preferences, particularly targeting under-25s who predominantly consume news via smartphones and social media platforms like TikTok. Announced by CEO Deborah Turness, this initiative is part of a broader restructuring to counter challenges such as news avoidance, the rise of social platform news consumption, increased digital competition, and declining traditional broadcast viewership.
The department will focus on delivering personalized news experiences and unique user journeys while adhering to the BBC’s public service values to maintain trust, accuracy, and impartiality. Changes include replacing traditional newsroom structures to ensure content is optimized for all platforms, including traditional media, online, and the BBC News app, which has overtaken Apple News as the No. 1 UK news app. A new director will oversee the department, focusing on implementing and evaluating key investments and innovations.
Noxtua Secures $92M to Develop Sovereign AI for German Legal System
Noxtua, formerly known as Xayn, has raised $92.2 million (€81.2 million) in a Series B funding round to develop a sovereign AI tailored for the German legal system. The round was led by strategic investor C.H. Beck, Germany’s leading legal publisher, providing Noxtua access to an extensive archive of 55 million legal documents—the largest legal database in the German-speaking world.
Additional investors include Northern Data Group, CMS, and Dentons. Noxtua’s AI, developed from research at the University of Oxford and Imperial College London, is designed to research legal matters and draft documents in a legally compliant manner. The AI operates on Northern Data’s cloud infrastructure within Germany, ensuring data sovereignty. CEO Dr. Leif-Nissen Lundbæk emphasized the necessity of a German-trained model, as U.S.-based models often lack precision in European legal contexts.
A16z Backs UK Startup Dex to Revolutionize AI-Powered Recruitment
Andreessen Horowitz (a16z) has led a $3.1 million pre-seed funding round for Dex, a UK-based startup aiming to transform the recruitment landscape using AI-driven talent matchmaking. Founded by Paddy Lambros and Harry Uglow, former executives at venture firm Atomico, Dex positions itself as an “AI voice talent agent” that interacts with candidates to understand their skills, aspirations, and preferences.
Dex’s platform not only identifies suitable job opportunities but also assists candidates through the application process, interview preparation, and offer negotiations. The startup has already onboarded approximately two dozen UK tech firms, including two prominent unicorns, and plans a broader launch later this year. Other investors in this round include Concept Ventures, Meta board member Charlie Songhurst, Deliveroo COO Eric French, and Notion Capital partner Kamil Mieczakowski.
Grouphug Emerges from Stealth to Enhance WhatsApp Group Interactions with AI
Grouphug, a stealth-mode startup led by seasoned entrepreneur Felix Petersen, is developing an AI-powered platform to enrich WhatsApp group interactions. Currently, Grouphug offers a feature that transforms group chat texts into humorous memes, serving as a beta user acquisition strategy. However, the company’s broader vision involves leveraging generative AI to extract and deliver more value from the vast, untapped data within private group chats.
The startup has secured a €1.5 million ($1.7 million) pre-seed funding round led by Berlin-based Blueyard VC, with participation from Tiny VC, Meta board member Charles Songhurst, and Atlantic Labs. Petersen is joined by co-founders Joseph Djenandji, formerly of travel brand LostIn, and Matthew Balazsi, an AI and machine learning expert with a decade of experience.
Alibaba Launches Open-Source Qwen3 AI Model Surpassing OpenAI’s o1 and DeepSeek’s R1
Alibaba has unveiled Qwen3, a new series of open-source AI models that reportedly outperform OpenAI’s o1 and DeepSeek’s R1 on key benchmarks. The flagship model, Qwen3-235B-A22B, utilizes a mixture-of-experts architecture with 235 billion parameters, achieving superior results in areas like software engineering and mathematics, and approaching the performance of Google’s Gemini 2.5-Pro.
Qwen3 introduces a “Thinking Mode,” allowing users to toggle between rapid responses and more in-depth reasoning for complex queries. The models are accessible via platforms like Hugging Face, ModelScope, Kaggle, and GitHub, and are available under the Apache 2.0 open-source license. This release underscores China’s commitment to advancing AI capabilities and providing accessible, high-performance models to the global developer community.
Hugging Face Unveils $100 3D-Printed Robotic Arm to Democratize AI Robotics
Hugging Face, renowned for its AI development platform, has introduced the SO-101, a 3D-printed robotic arm priced from $100. Developed by its robotics division, LeRobot, in collaboration with French firm The Robot Studio, the SO-101 is an evolution of the earlier SO-100 model. Enhancements include improved motors for reduced friction and the capability to support its own weight. Equipped with a camera, the arm can be trained using reinforcement learning to perform tasks like picking up objects. While the base price is $100, fully assembled units may cost up to $500 due to assembly and import tariffs.
This initiative aligns with Hugging Face’s broader strategy to make AI and robotics more accessible. The company recently acquired French startup Pollen Robotics, known for its humanoid robot Reachy 2, further expanding its footprint in the AI robotics sector.
Huawei Developing New Ascend 910D AI Chip to Challenge Nvidia’s H100
Huawei, the Chinese technology giant, is reportedly making significant strides in developing a new artificial intelligence chip, the Ascend 910D, with the explicit goal of competing directly with Nvidia’s high-end H100 GPU. Reports indicate that Huawei is preparing to begin testing sample batches of the Ascend 910D with select Chinese tech companies as early as late May 2025. This development is largely seen as a strategic response by Huawei and China to ongoing U.S. export controls that restrict access to advanced foreign AI chips like the H100.
While Huawei’s previous AI chip, the Ascend 910C, was estimated to deliver around 60% of the H100’s inference performance, the new Ascend 910D is anticipated to significantly close this performance gap. The chip is reportedly being manufactured using SMIC’s 7nm N+2 process technology. Huawei’s push to develop a powerful domestic alternative highlights China’s ambition for self-sufficiency in AI hardware and the increasing competition in the global AI chip market. The success of the Ascend 910D could have a notable impact on Nvidia’s market share in China and the broader landscape of AI chip development.
OpenAI Adds Shopping Features to ChatGPT, Delivering Tailored Product Recommendations with Images and Reviews
OpenAI has upgraded ChatGPT’s search capabilities, introducing a suite of shopping features that provide users with personalized product recommendations, complete with images, reviews, prices, and direct purchase links. This update, available to all ChatGPT users-including those on Free, Plus, and Pro plans, as well as users not logged in-enables shoppers to ask highly specific questions in natural language and receive curated results across categories like fashion, electronics, beauty, and home goods.
OpenAI reports that ChatGPT handled over 1 billion web searches in the past week, underscoring the tool’s growing role in product discovery. The recommendations are organic-no ads or affiliate commissions-and are generated independently using structured data from external sources. Product cards display images, star ratings, and detailed descriptions, with a sidebar offering additional information and user reviews from platforms like Amazon and Reddit. OpenAI has also enhanced search verifiability with multiple source links and improved citation highlighting. This move positions ChatGPT as a user-centric alternative to Google’s ad-driven shopping experience, further intensifying the competition in the AI-powered search market.
Craif Secures $22 Million Series C to Scale Non-Invasive Early Cancer Detection in US and Beyond
Craif, a Japanese biotech startup spun out of Nagoya University, has raised $22 million in Series C funding to accelerate the global rollout of its AI-driven, urine-based early cancer detection platform. The round was led by X&KSK, with participation from Unreasonable Group, TAUNS Laboratories, Daiwa House Industry, and Aozora Bank Group, bringing Craif’s total funding to $57 million and valuing the company just under $100 million.
Craif’s flagship product, miSignal, uses microRNA (miRNA) biomarkers and advanced machine learning to detect cancer at early stages, including Stage 1, through a simple urine test. The test is already available in over 1,000 clinics and 600 pharmacies in Japan, serving around 20,000 users and generating $5 million in revenue for 2024, with a target of $15 million by year-end. Craif plans to expand miSignal’s coverage to more than ten cancer types and is also exploring applications for neurodegenerative diseases. The funding will support US expansion, with R&D operations in Irvine, California, and a new office planned for San Diego. Craif is partnering with 30 US medical institutions for clinical trials, aiming for FDA approval by 2027. The company’s approach stands out for its non-invasive methodology, high sensitivity, and cost-effectiveness, leveraging the scientific significance of miRNA, which gained further recognition after the 2024 Nobel Prize in Medicine.
Writer Launches Palmyra X5: A Cost-Effective Alternative to GPT-4
Enterprise AI company Writer has unveiled Palmyra X5, a large language model (LLM) offering near GPT-4.1 performance at 75% lower cost. The model features a 1-million-token context window, enabling it to process up to 1,500 pages of content in a single pass. Pricing is set at $0.60 per million input tokens and $6 per million output tokens, making it a cost-efficient alternative for enterprises.
Trained using synthetic data for approximately $1 million in GPU costs, Palmyra X5 is integrated with AWS Bedrock, simplifying deployment for developers. The model supports multi-step reasoning and function calling, facilitating the development of autonomous AI agents for business applications. Writer’s innovative training techniques and cost-effective approach position Palmyra X5 as a competitive option in the enterprise AI landscape.AI Week # 17 has proven one thing—innovation never sleeps. From record-breaking funding rounds to game-changing AI tools and powerful global collaborations, the future of artificial intelligence is unfolding at lightning speed. Stay ahead of the curve, stay informed. Don’t miss our next edition of AI Weekly as we bring you the most important updates from AI Week # 18—because in this era, every week shapes the future. Subscribe now, and be the first to know what’s next in AI.
