Every model, in full.

The open-weight coding models served on the Zro endpoint, with the numbers that matter: quantization, context, measured serving performance, and per-token pricing.

Get started Read the model docs

DeepSeek

DeepSeek V4.1 Flash

Fastest

Fast million-token reasoning model for coding, agents, and high-throughput workflows.

Specifications

Model ID: deepseek-v4.1-flash
Context window: 1M tokens
Max output: 64K tokens
Quantization: FP8 weights, FP4 indexer
Input: Text + Image
Features: Tools, Reasoning, JSON mode, Logprobs

From DeepSeek

Performance

Output speed: 220tok / sec
First token: 516ms
Cache hit rate (p50): 95.5%of prompt

Pricing

Input: $0.30
Output: $1.20
Cache read: $0.006

USD per 1M tokens

Z.ai

GLM-5.3

Most powerful

GLM 5.3 is Z.ai's flagship long-context reasoning model for complex coding and agentic workflows.

Specifications

Model ID: glm-5.3
Context window: 1M tokens
Max output: 131K tokens
Quantization: FP8
Input: Text
Features: Tools, Reasoning

From Z.ai

Performance

Output speed: 125tok / sec
First token: 700ms
Cache hit rate (p50): 97.6%of prompt

Pricing

Input: $1.40
Output: $4.40
Cache read: $0.26

USD per 1M tokens

Z.ai

GLM-5.3 Flash

Fast, million-token reasoning model for coding agents and tool-heavy workflows.

Specifications

Model ID: glm-5.3-flash
Context window: 1M tokens
Max output: 64K tokens
Quantization: FP8
Input: Text
Features: Tools, Reasoning

From Z.ai

Performance

Output speed: 236tok / sec
First token: 357ms
Cache hit rate (p50): 95.1%of prompt

Pricing

Input: $0.15
Output: $0.50
Cache read: $0.03

USD per 1M tokens

Z.ai

Dolly 1 Security

Uncensored

Uncensored security variant of GLM-5.3 Flash for defensive and red-team workflows.

Specifications

Model ID: dolly1-security
Context window: 1M tokens
Max output: 64K tokens
Quantization: FP8
Input: Text
Features: Tools, Reasoning

Performance

Output speed: 215tok / sec
First token: 328ms
Cache hit rate (p50): 96.1%of prompt

Pricing

Input: $0.45
Output: $1.50
Cache read: $0.09

USD per 1M tokens

Point your coding agent at any of them

Create an API key