zai
GLM 5.3 Flash
GLM-5.3-Flash is Z.ai’s native multimodal coding model, featuring 320B total parameters, 18B activated parameters, and a 1M-token context window. Its efficient hybrid attention architecture supports visual coding, tool use, and end-to-end professional workflows across code, browsers, documents, and graphical interfaces.
reasoningtool useimplicit cachingvisionstructured outputtextimage
Overview
- Type
- language
- Context
- 1M
- Max output
- 131K
- Input / output
- text, image / text
API
Model identifier
zai/glm-5.3-flashProviders 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| baseten | Operational | 0.54s / 2.70s | 97.0 / 97.0 tok/s | 99.99% / 99.99% / 99.97% | $0.15/M | $0.50/M | $0.03/M / — | — | us | Yes | Yes |
| boundless | Operational | 0.44s / 2.76s | 267.0 / 291.0 tok/s | 100.00% / 100.00% / 81.16% | $0.12/M | $0.40/M | $0.02/M / — | — | — | No | No |
| deepinfra | Operational | 1.35s / 8.29s | 17.0 / 21.5 tok/s | 100.00% / 99.46% / 99.39% | $0.07/M | $0.25/M | $0.01/M / — | — | — | Yes | Yes |
| digitalocean | Operational | 1.03s / 2.60s | 78.0 / 100.0 tok/s | 100.00% / 99.99% / 99.99% | $0.15/M | $0.50/M | $0.03/M / — | — | — | Yes | Yes |
| fireworks | Operational | 1.14s / 5.97s | 175.0 / 227.0 tok/s | 100.00% / 100.00% / 100.00% | $0.15/M | $0.50/M | $0.03/M / — | — | us | Yes | Yes |
| friendli | Operational | 0.23s / 0.86s | 195.0 / 231.9 tok/s | 100.00% / 95.09% / 97.09% | $0.15/M | $0.50/M | $0.03/M / — | — | — | No | No |
| gmicloud | Operational | 3.67s / 9.67s | 110.5 / 159.8 tok/s | 100.00% / 100.00% / 99.77% | $0.09/M | $0.30/M | $0.02/M / — | — | — | No | No |
| inference-net | Operational | 1.13s / 13.95s | 67.0 / 212.8 tok/s | 100.00% / 99.88% / 99.96% | $0.10/M | $0.45/M | $0.03/M / — | — | — | No | No |
| modal | Operational | 0.39s / 0.58s | 155.0 / 183.8 tok/s | 100.00% / 100.00% / 100.00% | $0.45/M | $1.50/M | $0.09/M / — | — | — | Yes | Yes |
| morph | Operational | 2.15s / 9.21s | 15.5 / 107.3 tok/s | 100.00% / 100.00% / 99.35% | $0.04/M | $0.65/M | $0.02/M / — | — | — | Yes | Yes |
| novita | Operational | 1.18s / 2.88s | 93.0 / 256.5 tok/s | 100.00% / 100.00% / 100.00% | $0.09/M | $0.30/M | $0.02/M / — | — | — | — | No |
| parasail | Operational | 0.49s / 1.11s | 322.5 / 461.0 tok/s | 100.00% / 100.00% / 99.84% | $0.15/M | $0.50/M | $0.03/M / — | — | — | Yes | Yes |
| particle | Operational | 3.71s / 11.74s | 57.0 / 113.0 tok/s | 100.00% / 100.00% / 100.00% | $0.10/M | $0.40/M | $0.03/M / — | — | — | Yes | Yes |
| relace | Operational | 1.85s / 6.15s | 56.5 / 95.9 tok/s | 100.00% / 100.00% / 99.89% | $0.01/M | $0.50/M | $0.01/M / — | — | — | Yes | Yes |
| runinfra | Operational | 0.57s / 2.91s | 517.5 / 531.9 tok/s | 100.00% / 100.00% / 100.00% | $0.11/M | $0.45/M | $0.03/M / — | — | — | No | No |
| runware | Operational | 0.69s / 1.61s | 114.0 / 114.0 tok/s | 100.00% / 100.00% / 99.99% | $0.15/M | $0.50/M | $0.03/M / — | — | — | Yes | Yes |
| streamlake | Operational | 1.14s / 3.28s | 108.0 / 127.5 tok/s | 100.00% / 100.00% / 99.72% | $0.15/M | $0.50/M | $0.03/M / — | — | — | No | No |
| togetherai | Operational | 1.52s / 7.40s | 222.0 / 361.0 tok/s | 100.00% / 100.00% / 99.99% | $0.15/M | $0.50/M | $0.03/M / — | — | — | Yes | Yes |
| wafer | Operational | 0.74s / 1.87s | 194.0 / 326.0 tok/s | 100.00% / 100.00% / 99.99% | $0.12/M | $0.50/M | $0.11/M / — | — | — | Yes | Yes |
| zai | Operational | 2.22s / 3.36s | 95.0 / 152.6 tok/s | 100.00% / 100.00% / 100.00% | $0.15/M | $0.50/M | $0.03/M / — | — | — | — | No |
Pricing
input
0.00000015
output
0.0000005
input cache read
0.00000003
varies by provider
Yes
regional
{
"us": {
"input": "0.00000015",
"output": "0.0000005",
"input_cache_read": "0.00000003"
}
}
Capabilities
reasoningtool useimplicit cachingvisionstructured outputtextimage
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
131K
released
Aug 26, 2026
temperature control
Yes
zero data retention
some
no training
some
regions
us
api specifications
v2, v3, v4