GLM 5.2 cut across 3 hosts, by up to 72% at Decart (mxfp4) (input)
Published 30 September 2026
Price move, per 1M tokens
- Decart (mxfp4) · input↓ 72%$1.40 → $0.39
- Decart (mxfp4) · output↓ 45%$4.40 → $2.40
- Decart (mxfp4) · cache read↓ 42%$0.26 → $0.15
- Relace · input↓ 33%$0.200 → $0.135
- Relace · outputunchanged$4.00 → $4.00
- Relace · cache read↓ 33%$0.200 → $0.135
- Reka · inputunchanged$1.40 → $1.40
- Reka · outputunchanged$4.40 → $4.40
- Reka · cache read↓ 23%$0.26 → $0.20
| host | rate | was | now | change |
|---|---|---|---|---|
| Decart (mxfp4) | input | $1.40 | $0.39 | ↓ 72% |
| Decart (mxfp4) | output | $4.40 | $2.40 | ↓ 45% |
| Decart (mxfp4) | cache read | $0.26 | $0.15 | ↓ 42% |
| Relace | input | $0.200 | $0.135 | ↓ 33% |
| Relace | output | $4.00 | $4.00 | — |
| Relace | cache read | $0.200 | $0.135 | ↓ 33% |
| Reka | input | $1.40 | $1.40 | — |
| Reka | output | $4.40 | $4.40 | — |
| Reka | cache read | $0.26 | $0.20 | ↓ 23% |
Decart (mxfp4) input: ↓ 72%Relace input and cache read: ↓ 33%Reka cache read: ↓ 23%
If you buy
Reka's cache-read cut makes long-context work with heavy prompt reuse easier to justify here, so re-check your cached-token share before your next top-up.
The model
GLM 5.2open the model →
Open weights753.3B parameters1049K context34 hosts
intelligence23rd of 168 writing21st of 168 coding37th of 168 agents14th of 55
Available from 30 hosted offers, this large downloadable model from Zhipu carries a permissive MIT licence and a one-million-token request limit. It is a strong choice for teams that want top-tier capability without licensing restrictions.
Source
Read on openrouter.ai · we published this on 30 September 2026.
read the original at openrouter.ai ↗