Qwen3.8 27B repriced across 4 hosts: Wafer input and cache read down 65%, Ionstream input up 63%
Published 29 September 2026
Price move, per 1M tokens
- Wafer · input↓ 65%$0.072 → $0.025
- Wafer · outputunchanged$4.40 → $4.40
- Wafer · cache read↓ 65%$0.057 → $0.020
- Reka · input↓ 65%$0.072 → $0.025
- Reka · outputunchanged$4.40 → $4.40
- Reka · cache read↓ 65%$0.057 → $0.020
- Ionstream · input↑ 63%$0.089 → $0.145
- Ionstream · outputunchanged$2.55 → $2.55
- Ionstream · cache readunchanged$0.085 → $0.085
- Phala · input↓ 17%$0.24 → $0.20
- Phala · output↓ 0.0%$2.5 → $2.5
- Phala · cache read↓ 0.0%$0.050 → $0.050
| host | rate | was | now | change |
|---|---|---|---|---|
| Wafer | input | $0.072 | $0.025 | ↓ 65% |
| Wafer | output | $4.40 | $4.40 | — |
| Wafer | cache read | $0.057 | $0.020 | ↓ 65% |
| Reka | input | $0.072 | $0.025 | ↓ 65% |
| Reka | output | $4.40 | $4.40 | — |
| Reka | cache read | $0.057 | $0.020 | ↓ 65% |
| Ionstream | input | $0.089 | $0.145 | ↑ 63% |
| Ionstream | output | $2.55 | $2.55 | — |
| Ionstream | cache read | $0.085 | $0.085 | — |
| Phala | input | $0.24 | $0.20 | ↓ 17% |
| Phala | output | $2.5 | $2.5 | ↓ 0.0% |
| Phala | cache read | $0.050 | $0.050 | ↓ 0.0% |
Wafer input and cache read: ↓ 65%Reka input and cache read: ↓ 65%Ionstream input: ↑ 63%Phala input: ↓ 17%
If you buy
If you run this model on Ionstream, your input spend just rose, so re-check your usage mix and cache settings before the next top-up.
The model
Qwen3.8 27Bopen the model →
Open weights27.8B parameters262K context20 hosts
intelligence63rd of 168 writing93rd of 168 coding45th of 168 agents33rd of 55
Qwen3.8 27B is a downloadable model you can run yourself, and its measured strength is maths and reasoning. Agentic recovery and tool use sit near the bottom of the field, so it suits a solver rather than an agent that has to get itself back on track.
Source
Read on openrouter.ai · we published this on 29 September 2026.
read the original at openrouter.ai ↗