GLM 5.3 Flash Cyber

glm-5.3-flash-cyber

GLM 5.3 Flash tuned for security work: code review, exploit analysis and incident response. A million tokens of context, tool calling and vision. Security-tuned, not an uncensored model.

Modalities

Price per 1M

$0.30 / $1.00

input / output, $0.15 cached input

Context

1.05M

1,048,576 tokens

The GLM 5.3 Flash Cyber API

Wild West API serves GLM 5.3 Flash Cyber under the model id glm-5.3-flash-cyber. It is GLM 5.3 Flash tuned for security work: code review, exploit analysis and incident response. It has 1.05M tokens of context, costs $0.30 per million input tokens and $1.00 per million output tokens, with cached input at $0.15, and supports tool calling and reads images. One key works on the OpenAI-compatible endpoint (https://wildwestapi.com/v1) and on the Anthropic-compatible one Claude Code uses.

It always reasons before it answers. A request that asks it not to still works, and the answer comes back with its reasoning, which is billed as output.

Security-tuned, not uncensored

GLM 5.3 Flash Cyber has not had its refusals taken out, so it carries no uncensored tag and it can decline. If it turns down a task you need done, Hipshooter, the uncensored GLM 5.3 Flash, is a separate option.

GLM 5.3 Flash Cyber or Apex Flash?

Apex Flash is tuned for pentesting, CTFs and malware analysis, and GLM 5.3 Flash Cyber for code review, exploit analysis and incident response. Both are priced the same. They take the same calls, so changing one string runs the same prompt on the other, which is an easy way to compare drafts. See Apex Flash.

Playbooks to start from

These are written around Apex Flash. To run one here, put this id in the model field and keep the prompt.

Run a security review on a pull request with Apex FlashAI security code review: send a git diff to Apex Flash and get findings with severity, line references and fixes. OWASP-aligned prompt and curl call. Build an incident timeline from raw logsAI incident timeline from logs: normalise timestamps, order events and flag gaps across auth and web logs with Apex Flash, then verify every row against source. Map security alerts to MITRE ATT&CK techniquesAI MITRE ATT&CK mapping: map alert text to technique ids with evidence and confidence using Apex Flash, then verify every id against attack.mitre.org. Turn a threat report into draft Sigma rulesAI for writing Sigma rules: feed a threat report paragraph to Apex Flash, get draft detection rules with logsource and ATT&CK tags, then test them properly. Threat model a new feature with STRIDE and Apex FlashAI for STRIDE threat modelling: describe a feature and data flow, get threats per element and mitigations to review with your team. Prompt and call included. Draft a tabletop exercise scenario with injectsAI for tabletop exercise scenarios: draft a ransomware outage scenario with timed injects and discussion questions with Apex Flash, then tailor it to your team.

All twelve are on the cybersecurity playbooks page.

GLM 5.3 Flash Cyber against the rest of the list

Model	Kind	Context	Input	Output	Tools	Images
Slingshooter glm-5.3-outlaw	Uncensored	1.05M	$2.00	$5.00	Yes	No
Hipshooter glm-5.3-flash-outlaw	Uncensored	1.05M	$0.30	$1.00	Yes	Yes
DeepSeek V4.1 Flash 1017 Outlaw deepseek-v4.1-flash-1017-outlaw	Abliterated	1.05M	$1.25	$2.50	Yes	Yes
DeepSeek V4.1 Flash Lite Outlaw deepseek-v4.1-flash-lite-outlaw	Abliterated	512K	$0.36	$1.44	Yes	Yes
High Noon mimo-v2.6-flash-outlaw	Abliterated	1.05M	$1.00	$2.00	Yes	Yes
Sidewinder qwen3.8-27b-outlaw	Uncensored	1.05M	$1.50	$5.00	Yes	Yes
Qwen3.8 27B Ultrafast Outlaw qwen3.8-27b-fast-outlaw	Uncensored	524K	$1.75	$5.50	Yes	Yes
Apex Flash apex-flash	Security-tuned	1.05M	$0.30	$1.00	Yes	Yes
GLM 5.3 Flash Cyber glm-5.3-flash-cyber	Security-tuned	1.05M	$0.30	$1.00	Yes	Yes

Prices are per million tokens. The same list answers at GET https://wildwestapi.com/v1/models.

Call it with curl

Shell

curl https://wildwestapi.com/v1/chat/completions \
  -H "Authorization: Bearer sk-ww-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3-flash-cyber",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Call it with the OpenAI SDK

Python

from openai import OpenAI

client = OpenAI(base_url="https://wildwestapi.com/v1", api_key="sk-ww-...")

resp = client.chat.completions.create(
    model="glm-5.3-flash-cyber",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

In Claude Code

Claude Code adds /v1/messages itself, so its base URL has no path. The Claude Code setup guide has the rest.

Shell

export ANTHROPIC_BASE_URL=https://wildwestapi.com
export ANTHROPIC_AUTH_TOKEN=sk-ww-...
export ANTHROPIC_MODEL=glm-5.3-flash-cyber

FAQ

Is GLM 5.3 Flash Cyber an uncensored model?

No. It is tuned for security work and has not had its refusals taken out, so it can decline. Uncensored models refuse far less and can still decline; abliterated ones rarely refuse. Both are on the model list.

Can GLM 5.3 Flash Cyber call tools and read images?

Tool calling: yes. Images: yes. Its reasoning stays on, and comes back with the answer.

How is it billed?

By the tokens each request reads and writes, including any reasoning, from prepaid credit. There is no subscription. Pricing has the details.