SSyncDev

GPT-6 Cyber

SyncDev model directory · Profile updated Nov 17, 2026

Compare with another model

GPT-6 Cyber is OpenAI's restricted security model for authorised vulnerability investigation, reproduction and fix verification, for Daybreak Red customers.

Key specifications

Specification score
70.9
Context window
1.05M
Max output
128K
Input / 1M tokens
From $20.00
Output / 1M tokens
From $100.00
License
Proprietary

The specification score summarizes recorded features and limits. It is not a benchmark result.

Pricing conditions: Indicative, restricted access. Up to 272K input: $20 input / $2 cached / $100 output per 1M; cache writes $25 (30-min) or $50 (12-h). Over 272K: $40 / $4 / $150. Per-call tool fees extra. Confirm with your OpenAI account team.

Inputs and outputs

Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.

Accepts

  • Text input
  • Image input

Produces

  • Text output

Capabilities

ReasoningFunction callingVision

What is GPT-6 Cyber?

You cannot sign up for this one with a credit card. GPT-6 Cyber is gated behind OpenAI's Daybreak Red programme, and an organisation needs both Daybreak Red access and explicit approval confirmed by its account team before the first request will succeed. The model exists to investigate vulnerabilities, reproduce them in an authorised test environment and verify that fixes hold.

It launched on 5 November 2026 in preview, succeeding GPT-5.6 Cyber.

Who gets in

Access runs through the API and Codex for approved Daybreak Red customers, with a wider Daybreak Red rollout planned. The setup is organisational, not per developer: the organisation and project must be configured before calls go through. If you are evaluating vendors for a security team, start that paperwork early, because nothing on this page can be tested until it clears.

OpenAI rates the model's exploit capability as "critical" under its Preparedness Framework, a step up from "high" for the previous generation. That rating explains the gate.

Specs for approved teams

Field Value
API ID gpt-6-cyber (single snapshot)
Status Preview, restricted
Context window 1,050,000 tokens
Max input 922,000 tokens
Max output 128,000 tokens
Input / output Text and image in, text out
Endpoints Chat Completions, Responses
Reasoning tokens Yes

Batch is not supported. OpenAI does not state effort levels or a knowledge cutoff for it, so do not assume they mirror other GPT-6 models.

Through Responses the full hosted toolbox is available: function calling, web search, file search, tool search, code interpreter, hosted shell, apply_patch, skills, computer use, MCP and image generation. Hosted shell plus computer use is the combination that matters for reproducing a bug inside a sandbox.

Two context bands, two price lists

Up to 272K input tokens, the rate is $20 input and $100 output per million tokens, with cached input at $2. Cache writes cost $25 for 30-minute retention or $50 for 12-hour retention. Past 272K, input becomes $40, cached $4 and output $150. Tool calls such as search and computer use carry their own per-call fees on top.

Those numbers make it one of the most expensive text models OpenAI sells: twice GPT-6 Astra on input and output. Treat every figure as indicative and confirm against your contract, since restricted programmes often carry negotiated terms.

Reading the capability claims

OpenAI reports an ExploitGym score of 48.5%, against 32.1% for GPT-5.6 Cyber. One caveat: we could only confirm that score second-hand, not in OpenAI's developer documentation. The ID, context, endpoints and prices above are documented and solid; the benchmark is publisher-reported and unverified. Ask your account team for the system card before relying on it.

Moving off the 5.5 preview

Its launch came with a deprecation notice. gpt-5.5-cyber-preview shuts down on 1 February 2027, and OpenAI's suggested replacement is simply the most capable cyber model available to your organisation. For approved Daybreak Red customers, that is this one.

Questions to settle in a pilot

Pick three known, already-patched vulnerabilities from your own codebase. Give the model the vulnerable commit and a sandbox, and see whether it reproduces each one, then whether it confirms the patch closes it. Record tokens per finding, tool-call fees and how often a human had to correct its reasoning. A security model that is right but costs more than an analyst's afternoon per bug still needs a business case.

For general-purpose alternatives that do not require vetting, see the OpenAI lineup or the coding models directory.

How to evaluate GPT-6 Cyber

Compare the model on the work you need to complete.

Tasks and inputs

GPT-6 Cyber is listed for Text generation, Coding & Code Generation, Computer Use. It accepts text, image and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.

Capacity and cost

Documented limits: 1.05M context and 128K maximum output. The listed route has starting rates of $20.00 input and $100.00 output per million tokens. Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.

Choosing a deployment

Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.

Frequently asked questions

Only organisations with OpenAI Daybreak Red access and approval confirmed by their account team. It is available to them through the API and Codex.

These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.

Google

Gemini 4 Hydrogen

Gemini 4 Hydrogen is Google's fast, lower-cost Gemini 4 model for everyday coding and agentic execution, with a 1M context, xhigh thinking and a free tier.

1.05M context

particle-ai

Particle-1

Particle-1 is Particle's first software-engineering model, an RL-tuned DeepSeek V4.1 Flash that aims to solve coding tasks with fewer tokens.

1.05M context

Microsoft

MAI-Cyber-1.1-Flash

MAI-Cyber-1.1-Flash is Microsoft AI's vulnerability-discovery model, available only inside its MDASH scanning harness to verified security customers.

— context

recursive

Flint

Recursive Flint is an open-weight 39.8B MoE coding-agent model with 2.5B active parameters, free to try via OpenCode and OpenRouter.

262.14K context

Deepreinforce

Ornith-1.5 122B

Ornith-1.5 122B is an open MIT-licensed MoE coding and agent model with a 262K context, reasoning by default and single-80GB-GPU serving.

262.14K context

MiniMax

MiniMax M3.1 Flash

MiniMax M3.1 Flash is MiniMax's GA open-weight multimodal MoE for coding agents: 428B total, 23B active, 1M context, from $0.15 per 1M input tokens.

1M context