DeepSeek V4 Flash repriced across 5 hosts: AtlasCloud output down 65%, Inceptron output up 71%
Published 29 September 2026
Price move, per 1M tokens
- Inceptron · input↓ 14%$0.065 → $0.056
- Inceptron · output↑ 71%$0.38 → $0.65
- Inceptron · cache readunchanged$0.027 → $0.027
- AtlasCloud · input↓ 48%$0.1584 → $0.0826
- AtlasCloud · output↓ 65%$0.48 → $0.17
- AtlasCloud · cache read↑ 64%$0.01008 → $0.01652
- Venice · input↑ 27%$0.138 → $0.175
- Venice · output↑ 27%$0.275 → $0.350
- Venice · cache read↑ 25%$0.028 → $0.035
- Wafer · input↑ 19%$0.059 → $0.070
- Wafer · outputunchanged$0.35 → $0.35
- Wafer · cache readunchanged$0.053 → $0.053
- Relace · input↓ 14%$0.021 → $0.018
- Relace · outputunchanged$0.32 → $0.32
- Relace · cache read↑ 12%$0.016 → $0.018
| host | rate | was | now | change |
|---|---|---|---|---|
| Inceptron | input | $0.065 | $0.056 | ↓ 14% |
| Inceptron | output | $0.38 | $0.65 | ↑ 71% |
| Inceptron | cache read | $0.027 | $0.027 | — |
| AtlasCloud | input | $0.1584 | $0.0826 | ↓ 48% |
| AtlasCloud | output | $0.48 | $0.17 | ↓ 65% |
| AtlasCloud | cache read | $0.01008 | $0.01652 | ↑ 64% |
| Venice | input | $0.138 | $0.175 | ↑ 27% |
| Venice | output | $0.275 | $0.350 | ↑ 27% |
| Venice | cache read | $0.028 | $0.035 | ↑ 25% |
| Wafer | input | $0.059 | $0.070 | ↑ 19% |
| Wafer | output | $0.35 | $0.35 | — |
| Wafer | cache read | $0.053 | $0.053 | — |
| Relace | input | $0.021 | $0.018 | ↓ 14% |
| Relace | output | $0.32 | $0.32 | — |
| Relace | cache read | $0.016 | $0.018 | ↑ 12% |
Inceptron input: ↓ 14%Inceptron output: ↑ 71%AtlasCloud output: ↓ 65%AtlasCloud cache read: ↑ 64%Venice input and output: ↑ 27%Wafer input: ↑ 19%Relace input: ↓ 14%Relace cache read: ↑ 12%
If you buy
On Inceptron, output costs more while input fell slightly, so re-check your input-to-output mix before your next top-up and whether this host still suits your workload.
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 29 September 2026.
read the original at openrouter.ai ↗