A plain-English guide for New Zealand businesses, current as of July 2026.
AI isn’t a single product any more, it’s a marketplace. These days the useful question isn’t “should we use AI?” but “which AI, for which job?” OpenAI, Anthropic, Google, DeepSeek, xAI, Meta, Mistral and others each offer a whole family of models, and they’re not interchangeable.
Different engines for different jobs
Think of it like a vehicle fleet. You wouldn’t send a courier van to do a truck-and-trailer job, or a truck to drop off one parcel across town. AI models are the same:
- Built for depth. The flagship “frontier” models handle complex reasoning, difficult coding, legal and financial analysis, and long multi-step tasks that need to hold together for hours. They’re the most capable, and the most expensive to run.
- Built for balance. Mid-tier models handle most everyday work (drafting, summarising, analysis, simple coding) for a fraction of the cost, and quickly enough for interactive use.
- Built for speed and scale. Lightweight models are for high-volume, repetitive jobs like classifying, tagging, simple Q&A or processing thousands of records, where throughput and cost matter more than brilliance.
Speed, cost and quality: pick two
In day-to-day use the big trade-offs are speed, cost and quality, and you rarely get all three. The cost gap is huge. A flagship model can cost 20 to 50 times more per unit of work than a lightweight one from the same vendor. For a one-off strategic analysis that’s money well spent. For a chatbot answering thousands of routine questions a day, it would blow the budget for no real gain.
A few other things increasingly separate the vendors:
- Context window: how much the model can take in at once. Some can now read a million-plus words (a whole codebase, or years of contracts) in one go.
- Multimodality: whether it handles images, audio and video, not just text.
- Openness: some models can be downloaded and run on your own infrastructure for full control of your data, while others are cloud-only.
- Fit with your tools: how well it plugs into what you already use, like Microsoft 365 or Google Workspace.
For most businesses the smart move isn’t to pick one winner. It’s to match the right model to each task, and plenty of teams now use two or three vendors depending on the job.
The main engines at a glance
| Vendor | Model | Current version | When to use it |
|---|---|---|---|
| OpenAI | GPT-5.5 | 5.5 (5.5 Pro for maximum precision) | The highest-stakes reasoning, coding and multi-tool agent work, where quality justifies the price. |
| OpenAI | GPT-5.4 mini / nano | 5.4 | Quick, cheap, high-volume jobs like chatbots, classifying, summarising and routing. |
| OpenAI | o-series | o3 | Careful step-by-step reasoning for hard maths, logic and analysis. |
| Anthropic | Claude Opus | 4.8 | The deepest reasoning and strongest coding and agent work, for long, quality-critical tasks. |
| Anthropic | Claude Sonnet | 4.6 | The everyday workhorse for writing, analysis and coding. Close to flagship quality for much less. A sensible default. |
| Anthropic | Claude Haiku | 4.5 | The fastest, cheapest Claude, for high-volume and time-sensitive work. |
| Gemini Pro | 3.1 Pro | Flagship reasoning with a huge context window and true multimodality (text, images, audio, video). Tight Google Workspace and Cloud integration. | |
| Gemini Flash | 3.5 Flash | A fast, cost-effective default for apps and high-volume tasks. | |
| Gemini Flash-Lite | 3.1 Flash-Lite | The cheapest Gemini, for translation, classification and high-throughput work. | |
| DeepSeek | DeepSeek V4-Pro | V4-Pro | An open-weight flagship, strong at coding and maths, for a fraction of Western prices. Can be self-hosted (MIT licence). |
| DeepSeek | DeepSeek V4-Flash | V4-Flash | Ultra-low-cost, high-volume work where budget comes first. Open weights for on-premise use. |
| xAI | Grok | 4.3 | Good-value frontier chat with real-time access to X data and video input. |
| Meta | Llama 4 (Scout / Maverick) | 4.5 refresh | Open source and self-hostable. The choice when privacy, on-premise control or avoiding per-use costs matter most. |
| Mistral | Mistral Large / Medium | Large 3 / Medium 3.5 | Europe’s leading open-weight family. Strong in many languages, with data sovereignty options and cost-efficient deployment. |
The bottom line
There’s no single “best” AI. There’s only the best fit for a particular task, budget and level of risk. OpenAI, Anthropic and Google swap the top spot back and forth month to month. DeepSeek, Meta and Mistral compete hard on cost and control. And xAI has carved out a niche on value and real-time data. For most businesses, the win comes from understanding the trade-offs and using the right tool for each job, not betting everything on one.
A note on pace: this space moves incredibly fast, with new versions and price changes almost every week. Everything above was current in July 2026, and as this was written, new releases (Fable 5 and GPT-5.6) were expected within days. Always check the vendor’s own model and pricing pages before you commit to anything in production.
You can also download a copy of the reference guide.
Not sure which AI tools make sense for your business? Book a free health check, or call us on 0800 688 768.