Primary use stated or inferred cautiously from official documentation.
O
OPENAI / gpt-6-cyber
GPT-6 Cyber
A variant of GPT-6 Astra adapted for authorized security research. It takes text and images and returns text through Responses or Chat Completions. Since November 5, initial API and Codex access has been limited to selected, approved Daybreak Red organizations.
01 / SUMMARY
The essential
Maximum input capacity when published by the source.
Documented maximum generation limit.
Verified availability channels.
Primary functional family and verified specialties.
Declared terms for API access, model weights, or self-hosting.
02 / TECHNICAL SHEET
Limits and integration
| API ID | gpt-6-cyber |
|---|---|
| Model type | Code · Text and reasoning · Multimodal |
| Access model | Paid proprietary |
| License | Proprietary |
| Deployment | Hosted API |
| Release | 05 NOV 2026 |
| Knowledge cutoff | Not published |
| Entrance | Text · Image |
| Exit | Text |
| Context window | 1,050,000 context tokens; up to 922,000 input |
| maximum output | 128,000 output tokens |
| Reasoning | Reasoning tokens supported |
| Published tools | Features |
| Structured Outings | Supported |
| Batch processing | Not compatible |
| Prompt cache | Not published |
| Fine-tuning | Not compatible |
| Verified platforms | OpenAI API · Codex |
Vulnerability triage, patch review and validation in authorized labs. A useful test checks whether a finding can be reproduced, whether the patch addresses its cause, and whether the model stays within the permitted assets and actions.
Access to GPT-6 Astra does not grant access to GPT-6 Cyber. The organization must meet Daybreak Red controls and define an authorized scope. Total context differs from the input limit: the documentation specifies 922,000 input and 128,000 output tokens. Batch and fine-tuning are unsupported.
03 / CAPABILITIES
What it can do
Orientation
Vulnerability triage, patch review and validation in authorized labs. A useful test checks whether a finding can be reproduced, whether the patch addresses its cause, and whether the model stays within the permitted assets and actions.
Context
1,050,000 context tokens; up to 922,000 input · 128,000 output tokens
Tools and integration
Features
Access
Daybreak Red · approval required
Input modalities
TextImage
04 / PRICES
Documented cost
| Concept | Worth | Unit/condition |
|---|---|---|
| Standard input | 20 USD / 1M input tokens | Standard API rate |
| Cached input | 2 USD / 1M cached input tokens | Reading reused prefixes |
| Cache write or storage | Not published | The condition varies by provider |
| Standard output | 100 USD / 1M output tokens | May include reasoning tokens |
| Batch input | Not published | Asynchronous processing |
| Batch output | Not published | Asynchronous processing |
Standard short-context rates per million tokens: USD 20 input, USD 2 cached input and USD 100 output. Cache writes cost USD 25 for 30 minutes or USD 50 for 12 hours. These figures do not budget every context length or include possible tool charges.
i Prices change and may depend on level, region or context length. Check the source before making a decision.
05 / EVALUATIONS
How to read the results
“
OpenAI reports 48.5% on ExploitGym and 40.3% vulnerability recall on its web test without source code. The protocols differ. The latter measures recall, not precision or general success. Inferama has not reproduced the results, which depend on the environment, tools and test rules.
| Benchmark | Result | Metric | Source |
|---|---|---|---|
| ExploitGym | 48,5 % | Score | View source ↗ |
| OpenAI Web Pentesting | 40,3 % | Score | View source ↗ |
Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.
06 / VERSIONS
Chronology
GPT-6 Cyber
Version added to Inferama's verified catalog.
07 / SOURCES