3 November 2026 · Site Methodology & Pricing Notes · 5 min read
I was pulling OpenAI's current pricing docs for an unrelated check today and ran into a service-tier table I hadn't seen before. Our own GPT-6 Astra API Cost Calculator has a dropdown with four options: standard, Fast, Batch/Flex, and GPT-5.6 Sol for comparison. OpenAI's actual pricing page lists five tiers for Astra. We'd never added the fifth one.
Three modes, same model, same weights: Standard at $10.00 input / $50.00 output per million tokens, Fast at $20.00/$100.00 (exactly 2x, for lower latency), and Batch/Flex at $5.00/$25.00 (half price, for asynchronous jobs that don't need an instant reply). That's what the calculator has priced since we built it, and all three numbers check out against OpenAI's docs today -- no correction needed there.
OpenAI's pricing table has a row we'd missed: Ultrafast, available only on GPT-6 Astra and GPT-6.1 Sol, billed at $60.00 input / $300.00 output per million tokens. That's not a typo-adjacent number next to Fast's $20/$100 -- it's a full 6x the standard rate, and 3x Fast mode's price, for the same model producing the same output.
| Nível | Insira $/1 milhão | Saída $/1 milhão | vs Standard |
|---|---|---|---|
| Batch/Flex | $5.00 | $25.00 | 0.5x |
| Padrão | $10.00 | $50.00 | 1x |
| Fast | $20.00 | $100.00 | 2x |
| Ultrafast | $60.00 | $300.00 | 6x |
Line up the cheapest and most expensive way to run the exact same GPT-6 Astra weights and the spread is 12x -- Ultrafast costs twelve times what Batch/Flex costs for identical input and output quality, with latency as the only thing you're actually buying.
When we built this calculator we priced the three tiers that led OpenAI's launch announcement for Astra. Ultrafast isn't in the headline sentence of the pricing page -- it's a row further down in the full service-tier table, next to Scale/Reserved capacity pricing that works differently (prepaid commitments, not a simple multiplier). We built from the summary instead of the full table and never went back to recheck it. That's exactly the kind of gap our own previous pricing audit e batch-deadline check were trying to catch going forward -- this one just hadn't come up yet.
The practical risk only hits one kind of visitor: someone who opens our calculator specifically to price Ultrafast mode for a low-latency agent workload. Without the option, the natural assumption is that Fast mode's 2x price is the ceiling OpenAI offers. It isn't. The real ceiling is 6x, and budgeting against the wrong number here isn't a rounding error -- it's a 3x miss on exactly the workload that cares most about getting the number right before it scales.
Added the Ultrafast tier to the calculator's model-tier dropdown and comparison table at $60.00/$300.00, updated the FAQ and page copy to describe all five tiers instead of three, and re-verified the Standard, Fast and Batch/Flex numbers against OpenAI's current docs in the same pass -- all three were still correct. Fixed and deployed today.
New to API service-tier pricing in general? Start with the guias gratuitos de custo de API.
Implemente você mesmo: DigitalOcean – crédito grátis de $ 200 ↗ · VPS Hostinger ↗
Checked directly against developers.openai.com/api/docs/pricing, 2026-11-03. Reference figures, not live quotes -- confirm current terms on OpenAI's own pricing page before relying on a tier price for production budgeting.