ToolsAugust 21, 2026via Vercel Blog
GPT-5.6 Sol is now 50% off a lower price
Why it matters
OpenAI's pricing move on Sol makes frontier-class models more accessible for coding agents and production workloads. Practitioners budgeting token spend this quarter need to recalibrate.
Key signals
- GPT-5.6 Sol: input $2.00/M → $4.00/M list (then 50% off = $2.00/M), output $10.00/M → $20.00/M list (then 50% off = $10.00/M) on Default tier
- Input cost drop: 20%; output cost drop: 33%
- Discount applies across all service tiers (Default, Flex, Priority/fast mode)
- Cached tokens, cache writes, long-context (>272K) requests move proportionally
- Discount valid through September 18, 2026
- Model ID unchanged; existing requests auto-bill at new price
- Integration: Vercel AI Gateway + coding-agents framework for agent deployments
- Applies to BYOK (bring-your-own-key) at user's own rate
The hook
GPT-5.6 Sol just got 50% cheaper. Input costs dropped 20%, output a third—and the discount stacks on an already-lower list price.
OpenAI lowered list pricing for , and the 50% AI Gateway now applies to the new, lower price through September 18. Input drops 20%, output drops a third. GPT-5.6 Soldiscount
The discount applies on every OpenAI service tier:
Rates are per million tokens for requests up to 272K tokens. Cached tokens…