Secure AI for Business Users.
Fixed costs. No surprises.

The VC subsidized tokens era is coming to an end. Anthropic recently moved enterprise customers from flat rate to predominantly usage based pricing, leading to headlines such as: “Anthropic Hiked the Price for Power Users Amid Complaints Its Model Is Getting Worse: Its most dedicated customers could pay significantly more under a new usage-based pricing model.” Other articles followed such as, “GitHub Copilot’s price shakeup could end cheap AI coding as we know it: We’re seeing the beginning of the end for flat-rate AI plans, starting with GitHub switching to usage-based pricing for its Copilot AI plans.”

But this should not be a surprise. We always knew that AI model providers were unprofitable - as proven by the eye-watering amounts they were raising. As they are reaching the limits of the private markets, and getting ready to face the rigors of the public markets, they need to focus on profitability.

At the same time, open source / open weight models are getting better and better - by some estimates only running three months or so behind the frontier models. So why would businesses pay more for frontier models? Note that for frontier models, these business customers are essentially paying a premium to offset the massive model training costs as well as the enormous amounts these companies spend on free tiers to attract new customers. 

Data security is another factor concerning business customers. In recent months there have been incidents with headlines such as, “Claude’s code: Anthropic leaks source code for AI software engineering tool” and “Anthropic’s most dangerous AI model just fell into the wrong hands.” Last year there were serious concerns about “A Federal Judge Ordered OpenAI to Stop Deleting Data. Here's How That Could Impact Users' Privacy.” Even though frontier model vendors promise to not train on business customer data (if the right options are selected), there is always a risk that information will be logged by an agent somewhere and inadvertently released. Moreover, these frontier model sites are the top targets for hackers while individual business servers, especially ones that are not connected to the public internet can stay under the radar. 

So, why don’t businesses just use open source models on-prem or on private cloud servers? Because they would need to stitch together several different open source projects to deliver a complete solution, would need to make sure the models were configured to run optimally on the servers, and then would need to maintain the models as they evolved. This is not achievable for smaller businesses. This is especially true given the rapid pace of innovation in both language models and associated harnesses and tools that need to be used to deliver business value.

At Aible, we solved this problem - working closely with NVIDIA and other leading partners to certify Aible on desktop supercomputers, servers, and neoclouds so that the complete solution including the models can be installed with one click. We are also early adopters of and launch partners for the latest open source models such as the recently released Nemotron 3 Nano Omni. Aible runs every component of the generative AI solution - the models, the tools, the coordination natively on the same device so no data ever leaves the customers control. And, for the first 500 launch customers we are offering all of these capabilities for $20,000 / server for a year - no per user or per token cost. 

We also believe our customers need to experience the value of autonomous agents. As such, we are including unlimited “What’s Changed and Why” agents at the same flat rate. In the current turbulent times, customer behavior, logistics costs, counterparty risk, etc. are changing in ways we can’t anticipate. When we don’t know what questions to ask, traditional analytics can’t help. The Aible agents automatically analyze millions of patterns across your structured data to detect what's changed in business KPIs, why, and what to do about it — before market shifts hit the P&L. 

Aible includes leading open source models such as the NVIDIA Nemotron, Google Gemma, Meta Llama, and OpenAI GPT OSS model families. It also includes all of the tooling you need to use your structured, unstructured, and media data with these models. As models and tools change, all updates are included in the subscription. And you can buy this solution today from your existing partner including SHI, CDW, WWT, or any others supported by TD SYNNEX or our OEM partners.