Host Wafer (mxfp4) raised DeepSeek V4 Flash cache-read pricing by 1438%
Published 22 October 2026
Price move, per 1M tokens
- Wafer (mxfp4) · inputunchanged$0.70 → $0.70
- Wafer (mxfp4) · outputunchanged$0.17 → $0.17
- Wafer (mxfp4) · cache read↑ 1438%$0.013 → $0.200
| host | rate | was | now | change |
|---|---|---|---|---|
| Wafer (mxfp4) | input | $0.70 | $0.70 | — |
| Wafer (mxfp4) | output | $0.17 | $0.17 | — |
| Wafer (mxfp4) | cache read | $0.013 | $0.200 | ↑ 1438% |
cache read: ↑ 1438%
If you buy
Re-check your cache-heavy workloads on this host, since prompt reuse now costs far more and may no longer suit long-context chat or agent loops.
The model
DeepSeek V4 Flashopen the model →
Open weights290.9B parameters1049K context32 hosts
intelligence71st of 184 writing62nd of 184 coding71st of 184 agents24th of 66
DeepSeek V4 Flash is a text model you can download and run yourself, or reach through one of many hosts, and it is built to be cheap at volume. Its measured quality is the catch: on the LiveBench boards it sits near the bottom of the field, so treat it as a cost choice rather than a quality one.
Source
Read on openrouter.ai · we published this on 22 October 2026.
read the original at openrouter.ai ↗