The GPT-6 Sol model card on developers.openai.com is the primary operator source for Alberta shops budgeting Astra-family coding agents — not a newsroom tease and not a community screenshot.
What OpenAI lists for GPT-6 Sol
Per the API model docs, GPT-6 Sol “is built for complex coding and agentic workflows.” Callers use gpt-6-sol. Reasoning effort supports none through max (medium default). The card states a 1,050,000-token context window, 128,000 max output tokens, and an April 20, 2026 knowledge cutoff.
Prompt caching is billed on the same card: cached input at 10% of uncached input ($0.20), cache writes at 1.25× uncached input ($2.50). For longer sessions, OpenAI’s prompt caching guide covers stable prefixes and cache keys — light reading for Alberta teams tuning agent loops, not a separate price invent.
“GPT-6 Sol is built for complex coding and agentic workflows.”OpenAI API — GPT-6 Sol model documentation
What it costs (API docs + Help Center delta)
These are USD rates from the Sol model docs. Do not quote them as CAD. CAD is on the invoice. Long-context multipliers apply when input exceeds 272K tokens — then 2× input and cache rates and 1.5× output bill for the full request. Batch and Flex are 50% of Standard; Fast mode is 2× the applicable rates on the API card.
| Line | Input / 1M | Output / 1M | Notes |
|---|---|---|---|
| gpt-6-sol | $2.00 | $10.00 | Cached $0.20 · writes $2.50 · 1.05M ctx |
| GPT-5.6 Sol (Help Center) | $4.00 | $20.00 | Cached $0.40 — sourced compare row |
| gpt-6-luna (strip) | $0.10 | $0.50 | Cached $0.01 · writes $0.125 · 1.05M ctx |
| Sol long context (>272K) | 2× | 1.5× | Applies to full request · cache rates 2× |
OpenAI’s ChatGPT Rate Card (Enterprise token-based pricing) places GPT-6 Sol at $2.00 / $0.20 / $10.00 beside GPT-5.6 Sol at $4.00 / $0.40 / $20.00 on the Work and Codex table — a clear sourced half-price list delta versus the prior Sol promotional band. Prefer that Help Center comparison over unsourced community “50% lower” posts.
Luna strip (same Astra family)
GPT-6 Luna (API gpt-6-luna) is OpenAI’s efficient tier for focused, high-volume tasks: $0.10 input / $0.50 output per 1M, cached input $0.01, cache writes $0.125, same 1.05M context / 128K max output shape, knowledge cutoff May 18, 2026 — per the Luna model card. Use Luna for high-volume classification and routine transforms; keep Sol for complex coding and agentic workflows.
What Alberta operators should do
For AI automation Alberta and AI consulting Alberta teams, put Sol on the coding-agent harness first: re-run your longest repo tasks at gpt-6-sol, watch cached-input share, and budget the >272K band before pasting a whole monorepo. Drop high-volume focused jobs to Luna so Sol spend stays on the hard paths.
Foreign hosted API is still not a private Alberta path. If prompts include client credentials, tender packages, or plant notes, pair this briefing with Opcelerate’s AI consulting and private AI security lanes — and keep reading The Super Intelligence Times for source-backed model cards. Quote USD as published; expect CAD on the invoice.
The guardrail
Half the GPT-5.6 Sol Help Center list is not permission to skip evaluation. Sol is a coding/agentic tier, not AGI. Watch Fast mode (2× on the API card) and long-context multipliers before you celebrate a cheap pilot. Keep regulated workloads off public paste endpoints.
gpt-6-sol, $2/$10 per 1M with $0.20 cached input, 1.05M context, Apr 20 2026 cutoff. Use Help Center Sol $2/$10 vs GPT-5.6 Sol $4/$20 as the sourced list delta. Route high-volume focused work to Luna ($0.10/$0.50). Foreign API ≠ private/on-prem — security review before client data ships.