OpenAI cut GPT-5.6 Sol input and output pricing by 50%
Published 18 September 2026
Price move, per 1M tokens
- OpenAI · input↓ 50%$4.00 → $2.00
- OpenAI · output↓ 50%$20.00 → $10.00
- OpenAI · cache read↓ 60%$0.50 → $0.20
- OpenAI · cache write↓ 60%$6.25 → $2.50
| host | rate | was | now | change |
|---|---|---|---|---|
| OpenAI | input | $4.00 | $2.00 | ↓ 50% |
| OpenAI | output | $20.00 | $10.00 | ↓ 50% |
| OpenAI | cache read | $0.50 | $0.20 | ↓ 60% |
| OpenAI | cache write | $6.25 | $2.50 | ↓ 60% |
input and output: ↓ 50%
If you buy
Workloads that were too costly to run at this model's full rate now fit a smaller budget, so re-check your usage caps and cache settings before your next top-up.
The model
GPT-5.6 Solopen the model →
Proprietary1050K context8 hosts
intelligence15th of 168 writing11th of 168 coding13th of 168 agents6th of 55
OpenAI's top-tier reasoning and coding model can handle a 1,050,000-token request. It is strongest when your team is already using OpenAI's tooling and can use the cheaper first-party tier.
Source
Read on openrouter.ai · we published this on 18 September 2026.
read the original at openrouter.ai ↗