Input / 1M$10.00
Output / 1M$50.00
Context1.05M
THE AI FIELD GUIDE / AUSTRALIAN TRADE BUSINESSES
AI rankings for your trade.
Compare models for the everyday office work. Click a model for details.
AI model rankings
What’s on your to-do list?Start with the work, then choose the model.
01
SOTA / FRONTIER
For complex reasoning and the harder problems.
S+Frontier
Input / 1M$4.00
Output / 1M$20.00
Context1M
SAdvanced
Input / 1M$2.00
Output / 1M$10.00
Context1.05M
02
WORKHORSE
Everyday office work, with a useful balance of cost and capability.
AVersatile
Input / 1M$2.00
Output / 1M$10.00
Context1M
Input / 1M$0.75
Output / 1M$3.75
Context1.05M
BEfficient
Input / 1M$0.30
Output / 1M$1.20
Context1.05M
03
LIGHTWEIGHT
Small, repeatable jobs with clear instructions.
CFocused
Input / 1M$0.10
Output / 1M$0.50
Context1M
Input / 1M$0.10
Output / 1M$0.50
Context1.05M
No models match this search.
Try a different name or return to the full shortlist.
An editorial starting point. These models haven’t been benchmarked by Web Blend on trade tasks. Read our method.
Putting AI to work in your business Workflow ideas & practical guidance
02 / FROM A MODEL TO A WORKFLOW
The real win is
getting it into your day.
A model is one piece of the puzzle. Your notes, price list, job software and a person checking the result are what make it useful.
Get this working in my businessPractical AI setup from Web Blend. Built around your existing tools.
A PRACTICAL STARTING POINTREVIEW BEFORE USE
01
Start with what happened.
Give the model approved job notes and the facts it needs.
02
Turn the notes into a draft.
Ask for a clear report, a customer update or a quote scope.
03
Check it. Then use it.
A person confirms the facts, prices and promises before anything goes out.
THE SMALL PRINT, IN PLAIN ENGLISH
Useful beats impressive.
01
A model isn’t an app.
API costs are separate from chat subscriptions. Files, tools and software connections change what you can actually do.
02
Capability isn’t task fit.
A lightweight model can be the right choice for a simple job. The task, cost and checking effort matter together.
03
Evidence comes first.
We separate catalogue facts, editorial opinion and future tests. See our sources and method
Sources and model notes
Metadata checked 2026-10-11. USD API prices per million tokens; subscriptions and conditional charges are separate. Tier placement is editorial.
GPT-6 Astra
A candidate for your most involved office work.
Tier S+. Input $10.00, output $50.00 per million tokens. Context: 1,050,000 tokens.
- Highest base token price in this shortlist; test whether the extra capability saves review time.
- Prompts over 272,000 tokens raise the full request to $20 input / $75 output per million.
- Trade-task performance has not been tested by Web Blend.
Claude Opus 5.5
For complicated briefs that need careful judgement.
Tier S+. Input $4.00, output $20.00 per million tokens. Context: 1,000,000 tokens.
- Routine customer updates may need a much cheaper model.
- Check every quantity, commitment and claim against supplied facts.
- Trade-task performance has not been tested by Web Blend.
GPT-6.1 Sol
A starting point for document-heavy office tasks.
Tier S. Input $2.00, output $10.00 per million tokens. Context: 1,050,000 tokens.
- Prompts over 272,000 tokens raise the full request to $4 input / $15 output per million.
- A bookkeeper must verify invoice fields, matches and exceptions.
- Trade-task performance has not been tested by Web Blend.
Claude Sonnet 5.5
A practical shortlist option for everyday drafting.
Tier A. Input $2.00, output $10.00 per million tokens. Context: 1,000,000 tokens.
- Give it the approved scope, tone and missing-information rules.
- Final prices and promises still need an owner's review.
- Trade-task performance has not been tested by Web Blend.
Gemini 3.8 Flash
Worth trying when job notes arrive in mixed formats.
Tier A. Input $0.75, output $3.75 per million tokens. Context: 1,048,576 tokens.
- Media, reasoning and tools can change the total API bill.
- Media understanding does not establish measurements or compliance.
- Trade-task performance has not been tested by Web Blend.
DeepSeek V4.1 Flash
A lower-cost candidate for repeatable office jobs.
Tier B. Input $0.30, output $1.20 per million tokens. Context: 1,048,576 tokens.
- Prices shown are the OpenRouter snapshot; the direct API has separate peak/off-peak prices.
- Check provider route and data handling before using customer records.
- Trade-task performance has not been tested by Web Blend.
Claude Haiku 5.5
Lightweight work can be the right place to start.
Tier C. Input $0.10, output $0.50 per million tokens. Context: 1,000,000 tokens.
- Prompts over 100,000 tokens raise the full request to $0.50 input / $2.50 output per million.
- Escalate ambiguity and exceptions for human review.
- C describes the lightweight class, not a failing grade; trade tasks remain untested.
GPT-6 Luna
A small-cost option for focused admin work.
Tier C. Input $0.10, output $0.50 per million tokens. Context: 1,050,000 tokens.
- Prompts over 272,000 tokens raise the full request to $0.20 input / $0.75 output per million.
- Use a fixed template and a clear route for missing or conflicting information.
- C describes the lightweight class, not a failing grade; trade tasks remain untested.