Enter platform
Two models, fine-tuned for cybersecurity.
Understand the adversary. Build a stronger defense.
Frontier-scale reasoning.
Focused on the work that matters.
CyberKimi.
Deep context.
Operational clarity.
CyberKimi and CyberGLM are both fine-tuned for cybersecurity. Two specialized models for the code, the evidence, and the decisions that define security work.
Choose CyberKimi or CyberGLM for specialized cybersecurity reasoning, from security context to actionable analysis.
CONCEPTUAL WORKFLOW · NOT A LIVE SESSION
Model ablation and cyber tuning, informed by investigations, research, and operational playbooks.
No GPU request logs. No training on customer prompts. Sensitive analysis belongs to your team.
From domain tuning toward reinforcement learning in verifiable security environments.
DEFENSIVE INTELLIGENCE
Connect the evidence. Understand the behavior. Give your team a clearer path to response.
Translate suspicious behaviors into Sigma, YARA, and KQL.
Reconstruct the sequence. Plan containment and recovery.
Turn threat intelligence into testable hypotheses.
Review the authentication events from this incident. Identify the sequence of suspicious activity, explain the evidence, and propose a focused detection strategy.
DEFINE THE OUTPUT97 / 100 tasks solved
A completed, same-budget CyberPVP match. Three tasks reached a token or iteration limit.
The published campaign records the score progression from 2 to 16. ExploitBench assesses V8 d8; this result demonstrates renderer-equivalent V8 exploitation and does not claim a browser sandbox escape.
Sources: CyberPVP evidence repository · ExploitBench evidence package · Results published September 2026.
Bring specialized cybersecurity reasoning to your tools with CyberKimi or CyberGLM. See the documentation for model IDs and integration details.
# Connect to specialized cyber reasoning.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.adverserial.ai/v1",
api_key=os.environ["CYBERKIMI_API_KEY"]
)
result = client.chat.completions.create(
model="lordx64/cyberkimi",
messages=[{
"role": "user",
"content": "Help plan a threat hunt."
}]
)
print(result.choices[0].message.content)Included weekly credit, priority member support, and higher plan limits for serious security work.
A way in.
$20/ month
For focused exploration with included weekly credit.
≈ $22 of PAYG usage per average month
Member support
Make it a daily practice.
$50/ month
For regular security work with more included credit and higher limits.
≈ $65 of PAYG usage per average month
Priority member support
More room to operate.
$200/ month
For sustained security work with the highest included credit and plan limits.
≈ $260 of PAYG usage per average month
Priority member support
Included weekly credit, priority member support, and higher plan limits. Monthly equivalents use 52 weeks ÷ 12 months. Allowances are issued weekly; unused credits do not roll over.
USD / BILLED MONTHLYOne usage credit = $1 at CyberKimi PAYG rates: $4 / 1M uncached input, $0.40 / 1M cached input, and $20 / 1M output. Output, including reasoning, consumes credits and counts toward output limits. Daily and weekly limits both apply.
Membership includes priority member support and the limits shown on your plan. Daily limits reset at 00:00 UTC. Weekly limits reset every seven days from activation, with no rollover. CyberKimi supports up to 1M total context tokens, including input and output. Memberships cover CyberKimi and CyberGLM in chat and the API. Both models share the same credit, output and concurrency limits; usage consumes credits at each model’s PAYG rates.
After an included credit or output limit, wait for the reset or opt into paid usage. Top up $10, $25, $50, or $100 at public PAYG rates, with no member discount. Concurrency limits still apply. Membership allowances and paid wallet funds stay separate.
One prepaid wallet. No subscription required.
Pay only for the tokens you use.
Fine-tuned for cybersecurity.
Cybersecurity-focused. Actively being refined.
A new addition to the cybersecurity model lineup.
Access and pricing will be announced at launch.
Talk to us about your organization’s cybersecurity AI needs.
Specialized reasoning.
Built for the people who defend.
