While the general product news has slowed to nothing, there are a few still coming in the AI chatbot realm.
OpenAI has officially launched its real-time video capabilities for ChatGPT, nearly seven months after the initial announcement. This new feature, known as Advanced Voice Mode with vision, allows users to interact with ChatGPT by pointing their phones at objects and receiving near real-time responses. ChatGPT has introduced a new Projects feature aimed at organizing user interactions with the AI chatbot. This feature allows users to create themed bundles of conversations and relevant files, drastically reducing the clutter often associated with managing multiple chats.
Anthropic has launched its Claude 3.5 Haiku model, now accessible to all users through its Claude chatbot on web and mobile platforms. This model, which is smaller and faster, has been noted for outperforming larger models in key benchmarks while remaining competitively priced. According to Artificial Analysis, Claude 3.5 Haiku has a latency of 0.80 seconds for the first token and an output speed of 65.1 tokens per second. With a context window of 200,000 tokens, it surpasses OpenAI’s models, allowing for better handling of extensive inputs. While the service is free, users face a daily message limit, but can subscribe to the Claude Pro plan for twenty dollars a month to unlock increased usage and priority access. The model excels in tasks like processing large datasets and analyzing financial documents, making it a strong contender in the AI landscape as it competes with offerings from OpenAI and Google.
Google Cloud has launched Agentspace, a new service aimed at helping enterprises create and deploy artificial intelligence agents. Initially available through an early access program, Agentspace features a multimodal search agent that can answer complex questions, offer proactive suggestions, and take actions based on company data. Early adopter Deloitte is already using the platform to streamline knowledge management.
Why do we care?
These announcements are incremental, not groundbreaking. Real-time vision, while impressive, replicates existing capabilities in competing tools. Similarly, project organization features could be seen as catch-up functionality rather than innovation. That said, productivity generally happens in incremental.
MSPs should evaluate these tools for clients in data-heavy industries (e.g., legal, finance) where context window size and multimodal capabilities have direct ROI potential. And, The rapid-fire updates from OpenAI, Anthropic, and Google underscore the need for ongoing evaluation. Providers should focus on aligning tool capabilities with client-specific needs rather than buying into every new feature announcement.