OpenAI just dropped GPT-5.6 with three pricing tiers, and the tech Twitter is losing its mind. But if you run a 5-50 person business — HVAC, plumbing, electrical, construction, whatever — you probably don't care about the benchmark scores.
You care about one thing: "Which tier should I pay for, and what will it actually cost me per month?"
Let me cut through the noise.
The Three Tiers (SMB Translation)
| Tier | Pay-As-You-Go | Pro Subscription | Best For |
|---|---|---|---|
| Sol | $5 per 1M input tokens | $30/month | Complex one-offs: legal review, technical architecture, high-stakes decisions |
| Terra | $2.50 per 1M input tokens | $15/month | Daily work: emails, quotes, proposals, customer support, basic analysis |
| Luna | $1 per 1M input tokens | $6/month | High-volume repetitive work: data entry, simple Q&A, bulk processing |
The short answer: For most trades businesses, Terra at $15/month is the sweet spot. You get 90% of the value for 50% of the cost.
Real Cost Examples (Your Actual Business Tasks)
Let's talk numbers for work you actually do.
Task: Write a response to a customer inquiry
- Tokens used: ~2,000 input + ~500 output
- Luna cost: $0.002 (two-tenths of a cent)
- Terra cost: $0.005 (half a cent)
- Sol cost: $0.01 (one cent)
- Verdict: Luna or Terra. This is commodity work.
Task: Generate a detailed quote from project notes
- Tokens used: ~15,000 input + ~3,000 output
- Luna cost: $0.015
- Terra cost: $0.038
- Sol cost: $0.075
- Verdict: Terra. You want the extra reasoning for pricing accuracy.
Task: Review a contract or legal document
- Tokens used: ~50,000 input + ~5,000 output
- Luna cost: $0.05
- Terra cost: $0.125
- Sol cost: $0.25
- Verdict: Sol. This is high-stakes. Pay the premium.
Task: Analyze 6 months of invoices to find patterns
- Tokens used: ~200,000 input + ~10,000 output
- Luna cost: $0.20
- Terra cost: $0.50
- Verdict: Terra. Luna might miss nuances in financial data.
Monthly Cost Scenarios
Scenario 1: Light User (5-10 AI tasks per week)
- 40 customer emails
- 8 quotes/proposals
- 4 basic analyses
- Luna: ~$2/month
- Terra: ~$5/month
- Sol: ~$10/month
- Recommendation: Terra Pro ($15) — you're under the limit, flat rate is simpler
Scenario 2: Moderate User (daily AI work)
- 150 customer emails
- 30 quotes/proposals
- 10 analyses
- 5 contract reviews
- Luna: ~$8/month
- Terra: ~$20/month
- Sol: ~$40/month (if you use it for everything)
- Recommendation: Terra Pro ($15) + pay-as-you-go Sol for the 5 contract reviews (~$1.25) = ~$16.25/month
Scenario 3: Heavy User (AI-first workflows)
- 400 customer emails
- 80 quotes/proposals
- 40 analyses
- 20 contract reviews
- Luna: ~$25/month
- Terra: ~$60/month
- Sol: ~$120/month
- Recommendation: Terra Pro ($15) + heavy pay-as-you-go Sol usage (~$50) = ~$65/month
The Hybrid Strategy (What I Actually Recommend)
Don't pick one tier and stick with it. Use all three, strategically:
- Terra Pro ($15/month) — Your daily driver. Handles 90% of your work.
- Luna pay-as-you-go — For bulk processing: importing customer lists, cleaning up spreadsheets, generating simple reports.
- Sol pay-as-you-go — For the 5-10 high-stakes tasks per month: contract review, technical troubleshooting, strategic decisions.
Total monthly cost for a moderate user: ~$20-25
That's less than most SMBs spend on a single software subscription they barely use.
When NOT to Upgrade
Stick with Luna or Terra if you're doing:
- Customer service emails
- Basic content writing (social posts, blog drafts)
- Simple data extraction
- Scheduling and calendar management
- Standard quote generation
Upgrade to Sol (pay-as-you-go) for:
- Legal or compliance review
- Technical architecture decisions
- Financial analysis with high stakes
- Medical or safety-critical content
- Complex multi-step reasoning
The Bottom Line
GPT-5.6's tiered pricing is actually great for SMBs. You're not forced to pay Sol prices for Luna work.
My recommendation for a 5-50 person trades business:
- Start with Terra Pro at $15/month
- Add $5-10/month in pay-as-you-go for Sol when you need it
- Total: ~$20-25/month for AI that handles emails, quotes, analysis, and the occasional complex task
That's 2-3 hours of a skilled worker's time. If AI saves you more than that (and it should), it pays for itself.
Want to see what this looks like in practice? I've put together some automation templates that use this tiered approach — routing simple tasks to cheaper models and escalating only when needed. Check it out at smbscaleup.gumroad.com.
Pricing reflects GPT-5.6 launch announcements as of July 2026. Verify current rates on OpenAI's pricing page before making purchasing decisions.
Top comments (0)