The AI Models I Actually Use — And the Ones I'm Watching
I get asked which AI I use. The honest answer is more than one. I build websites and web applications for small businesses, and I test models regularly because the wrong model on the wrong task produces work I cannot ship. Here is what I have learned about who is good at what.
Claude (Anthropic) — the one I reach for when nuance matters
When the output represents me to a client, Claude is the default. It is distinctively good at reading between the lines — a client who goes quiet is not busy, they are cooling. A client who says "less corporate" is not giving copy notes, they are saying the work does not sound like them yet. That kind of emotional granularity matters when you are building something that represents another person's business.
Claude is also structurally complete. When I ask for a client expectation model with a gap table and risk flags, it delivers all four columns filled. It does not skip sections. For high-stakes work — architecture decisions, client-facing analysis, anything where missing a detail has real consequences — this is where I start.
DeepSeek — the one that matched Claude on structure
When I first tested DeepSeek on governance analysis — the structured reviews I run on every project before it ships — the output was structurally identical to Claude. Same gap tables. Same risk flags. Same emotional reading of client behavior. The difference was speed and cost: same quality, delivered faster and at a fraction of the cost.
That changed what is possible. Tasks that were previously budget-constrained — running full governance reviews mid-project instead of only at the beginning and end, for example — became routine. Most of my pipeline now runs here. Claude stays on the highest-stakes work where I want the extra margin. The pipeline is model-agnostic by design: swap the model, same output shape, pick the right one for the job.
OpenAI — the ecosystem player
OpenAI has the broadest integration surface. If a tool has built-in AI, it probably speaks OpenAI's API format first — including tools from other model providers. GPT-4o is a strong generalist with genuine multimodal capability: vision, voice, image understanding. It is rarely the best at any single thing, but it is almost never the worst, and the API format became the industry default. Worth knowing for that reason alone.
Google (Gemini) — the long-context specialist
Gemini ships with genuinely long context windows — a million tokens or more in production. For document-heavy work where the AI needs to hold an entire codebase or a large corpus in its attention at once, that capability is unique. Google ecosystem integration — Workspace, Cloud, Search — also matters if you are in that stack. I have not routed production tasks here yet, but the long-context capability is worth tracking for projects that involve large document sets.
Mistral — the sovereignty play
Mistral builds the strongest open-weight models available today. For clients with data-sovereignty requirements — regulated industries, European operations, anyone who needs the model running on their own infrastructure — self-hosted Mistral is the path. It is also fast: smaller models that punch above their weight class. Not in my production pipeline today, but the open-weight path is the long-term hedge against single-provider dependency.
Why not pick one and stick with it
Because picking one is a bet. And a bet is not a strategy.
Six months ago the model that was best at emotional granularity was not the model that is best at it today. The model that was cheapest was not the one that is cheapest now. The landscape shifts quarterly. If your workflow is hard-wired to one provider, every shift is a migration project. If your pipeline is model-agnostic, every shift is a configuration change.
I use Claude and DeepSeek day to day. I track OpenAI, Google, and Mistral because the right model for a given task today may not be the right model six months from now. The work does not change — the tool that does it best does.
If you are thinking about a new website or online store, you do not need to care about any of this. That is the point. Tell me about your business and I will handle the rest.