zai
GLM 5.3
GLM 5.3 delivers comprehensive advancements in complex software engineering and agent capabilities. It uses the same base model as GLM-5.2, with all improvements driven by post-training.
implicit cachingreasoningtool usestructured outputexplicit cachingtext
Overview
- Type
- language
- Context
- 1M
- Max output
- 1M
- Input / output
- text / text
API
Model identifier
zai/glm-5.3Providers and live performance
Rolling Gateway metrics. Latency is time to first token; throughput is output tokens per second.
| Provider | Status | Latency | Throughput | Uptime 15m / 1h / 1d | Input | Output | Cache read / write | Discount | Regions | ZDR | No training |
|---|---|---|---|---|---|---|---|---|---|---|---|
| alibaba | Operational | 1.27s / 8.53s | 115.5 / 160.6 tok/s | 100.00% / 100.00% / 100.00% | $1.40/M | $4.40/M | $0.28/M / — | — | — | Yes | Yes |
| baseten | Operational | 2.27s / 10.74s | 130.0 / 147.0 tok/s | 99.98% / 99.98% / 99.92% | $1.40/M | $4.40/M | $0.14/M / — | — | us | Yes | Yes |
| bedrock | Operational | 4.52s / 32.01s | 133.0 / 227.0 tok/s | 100.00% / 99.94% / 95.16% | $1.68/M | $5.28/M | $0.31/M / $2.10/M | — | us | Yes | Yes |
| boundless | Operational | 0.61s / 2.90s | 125.5 / 166.8 tok/s | 100.00% / 100.00% / 100.00% | $1.12/M | $3.52/M | $0.14/M / — | — | — | No | No |
| crusoe | Operational | 3.55s / 3.55s | — | 33.33% / 15.79% / 89.68% | $1.40/M | $4.40/M | $0.26/M / — | — | — | No | No |
| deepinfra | Operational | 0.44s / 2.69s | 128.0 / 205.0 tok/s | 100.00% / 100.00% / 99.24% | $0.56/M | $2.50/M | $0.13/M / — | — | — | Yes | Yes |
| digitalocean | Operational | 3.21s / 26.35s | 17.0 / 44.8 tok/s | 100.00% / 99.07% / 98.87% | $1.40/M | $4.40/M | $0.26/M / — | — | — | Yes | Yes |
| fireworks | Operational | 1.37s / 4.96s | 84.0 / 123.0 tok/s | 99.96% / 99.82% / 99.91% | $1.40/M | $4.40/M | $0.26/M / — | — | us | Yes | Yes |
| friendli | Operational | 0.20s / 0.36s | 126.5 / 178.4 tok/s | 100.00% / 100.00% / 100.00% | $1.26/M | $3.96/M | $0.23/M / — | — | — | No | No |
| gmicloud | Operational | 1.85s / 2.81s | 156.5 / 159.7 tok/s | 100.00% / 100.00% / 99.94% | $1.40/M | $4.40/M | $0.26/M / — | — | — | No | No |
| inceptron | Operational | 1.43s / 2.18s | 209.0 / 465.0 tok/s | 100.00% / 100.00% / 100.00% | $1.25/M | $4.40/M | $0.26/M / — | — | — | No | Yes |
| inference-net | Operational | 0.36s / 1.17s | 147.5 / 334.9 tok/s | 100.00% / 100.00% / 100.00% | $0.90/M | $3.00/M | $0.15/M / — | — | — | No | No |
| modal | Operational | 0.41s / 0.84s | 145.0 / 162.3 tok/s | 100.00% / 100.00% / 99.98% | $1.40/M | $4.40/M | $0.26/M / — | — | — | Yes | Yes |
| morph | Operational | 0.56s / 1.74s | 100.0 / 145.5 tok/s | 100.00% / 100.00% / 99.91% | $0.10/M | $6.00/M | $0.10/M / — | — | — | Yes | Yes |
| novita | Operational | 1.22s / 3.16s | 225.0 / 245.1 tok/s | 100.00% / 100.00% / 100.00% | $0.91/M | $2.86/M | $0.17/M / — | — | — | — | No |
| parasail | Operational | 0.56s / 3.34s | 191.0 / 264.0 tok/s | 100.00% / 100.00% / 98.98% | $1.40/M | $4.40/M | $0.26/M / — | — | — | Yes | Yes |
| relace | Operational | 1.14s / 1.46s | 116.0 / 128.3 tok/s | 100.00% / 100.00% / 99.77% | $0.06/M | $4.40/M | $0.06/M / — | — | — | Yes | Yes |
| runware | Operational | 11.73s / 33.14s | 24.0 / 106.0 tok/s | 100.00% / 99.74% / 99.86% | $1.20/M | $4.00/M | $0.20/M / — | — | — | Yes | Yes |
| streamlake | Operational | 1.07s / 1.90s | 184.5 / 211.8 tok/s | 100.00% / 100.00% / 99.94% | $1.40/M | $4.40/M | $0.26/M / — | — | — | No | No |
| togetherai | Operational | 0.75s / 5.60s | 94.0 / 155.5 tok/s | 100.00% / 100.00% / 99.97% | $1.40/M | $4.40/M | $0.26/M / — | — | — | Yes | Yes |
| wafer | Operational | 0.98s / 1.33s | 112.0 / 124.4 tok/s | 100.00% / 100.00% / 99.99% | $0.17/M | $4.80/M | $0.17/M / — | — | — | Yes | Yes |
| zai | Operational | 1.26s / 2.13s | 168.5 / 196.1 tok/s | 100.00% / 100.00% / 100.00% | $1.40/M | $4.40/M | $0.26/M / — | — | — | — | No |
Pricing
input
0.0000014
output
0.0000044
input cache read
0.00000014
varies by provider
Yes
fast
{
"input": "0.0000021",
"output": "0.0000066",
"input_cache_read": "0.00000021"
}
regional
{
"us": {
"input": "0.0000014",
"output": "0.0000044",
"input_cache_read": "0.00000014"
}
}
Capabilities
implicit cachingreasoningtool usestructured outputexplicit cachingtext
Supported parameters
max tokens
Yes
temperature
Yes
stop
Yes
tools
Yes
tool choice
Yes
reasoning
Yes
include reasoning
Yes
response format
Yes
structured outputs
Yes
Reasoning options
options
{
"type": "effort",
"values": [
"low",
"high",
"max"
]
}
Metadata
type
language
context window
1M
max output
1M
released
Aug 18, 2026
temperature control
Yes
zero data retention
some
no training
some
regions
us
api specifications
v2, v3, v4