Qwen3.8 27B cut across 2 hosts, by up to 40% at DekaLLM (input)
Published 22 September 2026
Price move, per 1M tokens
- DekaLLM · input↓ 40%$0.20 → $0.12
- DekaLLM · outputunchanged$2.50 → $2.50
- DekaLLM · cache readunchanged$0.050 → $0.050
- Wafer · input↓ 40%$0.20 → $0.12
- Wafer · outputunchanged$2.50 → $2.50
- Wafer · cache readunchanged$0.050 → $0.050
| host | rate | was | now | change |
|---|---|---|---|---|
| DekaLLM | input | $0.20 | $0.12 | ↓ 40% |
| DekaLLM | output | $2.50 | $2.50 | — |
| DekaLLM | cache read | $0.050 | $0.050 | — |
| Wafer | input | $0.20 | $0.12 | ↓ 40% |
| Wafer | output | $2.50 | $2.50 | — |
| Wafer | cache read | $0.050 | $0.050 | — |
DekaLLM input: ↓ 40%Wafer input: ↓ 40%
If you buy
DekaLLM's input rate for this model now sits lower, so workloads that were borderline on cost are worth re-running, and any budget set against the old figure should be checked.
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 22 September 2026.
read the original at openrouter.ai ↗