Host Phala cut Qwen3.5-27B cache-read pricing by 80%
Published 30 September 2026
Price move, per 1M tokens
- Phala · inputunchanged$0.30 → $0.30
- Phala · outputunchanged$2.40 → $2.40
- Phala · cache read↓ 80%$0.150 → $0.030
| host | rate | was | now | change |
|---|---|---|---|---|
| Phala | input | $0.30 | $0.30 | — |
| Phala | output | $2.40 | $2.40 | — |
| Phala | cache read | $0.150 | $0.030 | ↓ 80% |
cache read: ↓ 80%
If you buy
If you lean on long prompts with reused context, this host now suits that pattern far better, so re-check your cache-read line before your next top-up.
The model
Qwen3.5-27Bopen the model →
Open weights27.8B parameters262K context9 hosts
intelligence92nd of 168 writing96th of 168 coding93rd of 168
Qwen3.5-27B is a 27.8-billion-parameter multimodal model from Alibaba with a permissive Apache licence and a 262,144-token request limit. Its measured coding skill sits well above its general text level, making it a natural fit for code-heavy workloads rather than pure creative writing.
Source
Read on openrouter.ai · we published this on 30 September 2026.
read the original at openrouter.ai ↗