Anthropic has launched Claude 3.7 Sonnet, its first hybrid reasoning AI model designed to tackle complex problems more efficiently than previous versions. The model excels in areas such as mathematics and coding, and is available starting Monday through the Claude app, Anthropic’s API, Amazon Bedrock, and Google Cloud’s Vertex AI. It costs three dollars per million input tokens and fifteen dollars per million output tokens, the same as its predecessor. In conjunction with the new model, Anthropic has introduced a coding tool called Claude Code, which functions as an active collaborator capable of editing files and managing code on GitHub. Claude 3.7 Sonnet has demonstrated significant improvements in coding, finance, and legal tasks, and offers a more current knowledge base with a cut-off date of October 2024.
ChatGPT claims it has surpassed four hundred million weekly active users, marking a substantial increase from two hundred million reported in August twenty twenty-four.
Google has announced that its AI coding tool, Gemini Code Assist, is now available for free to individual users worldwide. Launched in public preview, this tool aims to assist solo developers, students, hobbyists, and startups by providing coding help without the need to switch between different windows or sources. The tool supports thirty-eight programming languages and can be installed in popular development environments such as Visual Studio Code and JetBrains.
Perplexity has launched a new tool called Deep Research, designed to conduct in-depth research and analysis, providing detailed reports in response to user questions. This tool is currently available for free with a limit of five queries per day for non-subscribers and up to five hundred queries for Pro users. The introduction of Deep Research follows similar announcements from competitors, including OpenAI and Google. Perplexity claims their tool excels at various expert-level tasks, taking approximately two to four minutes to generate an answer by performing numerous searches and analyzing hundreds of sources. The company asserts that it outperforms rivals on a specific benchmark, achieving an accuracy score of twenty-one point one percent, although this is still lower than that of OpenAI’s Deep Research.
The Allen Institute for AI and Contextual AI has launched OLMoE, a compact artificial intelligence model that allows developers to explore AI applications on iOS devices. This model is compatible with the iPhone 15 Pro and newer models, as well as M-series iPads and Macs, with smaller versions planned for older devices shortly. OLMoE can process over 40 tokens per second, designed for efficient text processing while safeguarding user privacy, since all data stays on the device and is erased after each session. Featuring seven billion parameters, it utilizes only one billion per input token, enabling smooth operation on edge devices. Developers can access this open-source app and the language model to develop and test their AI applications locally.
And in Microsoft moves, Microsoft is preparing for the launch of OpenAI’s upcoming models, GPT-4.5 and GPT-5. Sources indicate that Microsoft is readying server capacity for these new models, with GPT-4.5 expected to launch within weeks and GPT-5 anticipated by late May.
Microsoft has also recently canceled several leases for artificial intelligence data center capacity in the United States, reflecting growing skepticism among investors regarding the substantial investments made by major technology companies in AI infrastructure. A report from brokerage TD Cowen indicates a potential oversupply of AI computing resources at Microsoft, raising concerns that companies may be overestimating long-term demand for AI.
Why do we care?
These AI stories raise critical strategic questions:
- Where’s the real revenue in AI? If OpenAI, Google, and Anthropic are all offering free or low-cost AI services, will businesses ever pay significant amounts for AI-driven solutions?
- Is Microsoft overextending in AI? If Azure-based AI infrastructure isn’t generating expected demand, Microsoft’s AI-first strategy could face challenges. Or are they smart to start pulling back first from these massive capital expenditures?
- Will AI tools replace traditional development? AI-assisted coding tools like Claude Code, Gemini Code Assist, and Copilot could change how developers work but not eliminate the need for skilled engineers.
AI hype is real, but sustainable business models are still in question. While AI tools are improving, IT service providers should focus on AI solutions with clear, long-term business value rather than just chasing the latest trends.
