DeepSeek V4 Flash repriced across 4 hosts: Relace output down 50%, OpenInference input up 250%
Published 24 September 2026
Price move, per 1M tokens
- OpenInference · input↑ 250%$0.040 → $0.140
- OpenInference · output↑ 40%$0.50 → $0.70
- OpenInference · cache read↑ 114%$0.014 → $0.030
- Relace · input↓ 25%$0.040 → $0.030
- Relace · output↓ 50%$0.64 → $0.32
- Relace · cache readunchanged$0.016 → $0.016
- Inceptron · input↓ 9%$0.091 → $0.083
- Inceptron · output↓ 32%$0.613 → $0.414
- Inceptron · cache readunchanged$0.060 → $0.060
- Wafer · input↓ 8%$0.087 → $0.080
- Wafer · outputunchanged$0.35 → $0.35
- Wafer · cache read↓ 75%$0.080 → $0.020
| host | rate | was | now | change |
|---|---|---|---|---|
| OpenInference | input | $0.040 | $0.140 | ↑ 250% |
| OpenInference | output | $0.50 | $0.70 | ↑ 40% |
| OpenInference | cache read | $0.014 | $0.030 | ↑ 114% |
| Relace | input | $0.040 | $0.030 | ↓ 25% |
| Relace | output | $0.64 | $0.32 | ↓ 50% |
| Relace | cache read | $0.016 | $0.016 | — |
| Inceptron | input | $0.091 | $0.083 | ↓ 9% |
| Inceptron | output | $0.613 | $0.414 | ↓ 32% |
| Inceptron | cache read | $0.060 | $0.060 | — |
| Wafer | input | $0.087 | $0.080 | ↓ 8% |
| Wafer | output | $0.35 | $0.35 | — |
| Wafer | cache read | $0.080 | $0.020 | ↓ 75% |
OpenInference input: ↑ 250%Relace output: ↓ 50%Inceptron output: ↓ 32%Wafer input: ↓ 8%
If you buy
OpenInference users now pay more for the same model, so re-check your usage mix and whether a lighter prompt or shorter context keeps the bill where you expected it.
The model
DeepSeek V4 Flashopen the model →
Open weights290.9B parameters1049K context33 hosts
intelligence65th of 168 writing57th of 168 coding64th of 168 agents20th of 55
DeepSeek V4 Flash is a downloadable text model with a permissive licence, built for long-document work where the bill matters more than peak quality. Its measured quality is weak on the LiveBench boards and mid-pack on the Arena boards, so treat it as a cheap workhorse rather than a quality leader.
Source
Read on openrouter.ai · we published this on 24 September 2026.
read the original at openrouter.ai ↗