GPT-5.6 Cyber

GA

Purpose-trained cybersecurity model in the GPT-5.6 family, for approved defenders doing authorized vulnerability research, exploit validation and security testing. ACCESS IS GATED: OpenAI states it 'requires separate approval and provisioning' via the Daybreak program, so the published rate is not self-serve. Priced in its own 'Cyber models' table on the pricing page (12.50 / 1.25 cached / 75.00 per 1M, cache writes 15.625) - a table check-price-drift.mjs does not read, because that script anchors on the Standard-tier tables. 400K context, 128K max output, Feb 16 2026 cutoff, text+image in / text out, reasoning tokens supported. The pricing page's long-context columns are '-' for this model, i.e. no long-context tier is published, unlike Sol. 'daybreak-red-latest' is an alias currently pointing here (and 'daybreak-blue-latest' to gpt-5.6-sol); the aliases are stated to re-point as new Daybreak models ship, with pricing following the underlying model. No release date published (null).

GPT-5.6 Cyber by OpenAI costs $12.50 per 1M input tokens and $75 per 1M output tokens ($28.13/1M blended), with a 400K (400.000-token) context window. It is generally available (GA), with a knowledge cutoff of 16 Feb 2026.

Last verified: 13 Aug 2026 · sourced from official provider documentation

Provider
OpenAI
Status
GA
Input price
$12.50 / 1M tokens
Output price
$75 / 1M tokens
Cached input
$1.25 / 1M tokens
Blended price
$28.13 / 1M tokens
Context window
400.000 tokens (400K)
Max output
128.000 tokens
Modality
text, image
Open weights
No — API only
Knowledge cutoff
2026-02-16
Released
API string
gpt-5.6-cyber

Source: OpenAI official documentation ↗

FAQ

GPT-5.6 Cyber — questions & answers

How much does GPT-5.6 Cyber cost?

GPT-5.6 Cyber is priced at $12.50 per 1M input tokens and $75 per 1M output tokens ($28.13/1M blended at a 3:1 input-to-output ratio), with cached input at $1.25 per 1M tokens.

What is the context window of GPT-5.6 Cyber?

GPT-5.6 Cyber has a 400.000-token context window (400K), with up to 128.000 output tokens per request.

Is GPT-5.6 Cyber deprecated?

No — GPT-5.6 Cyber is generally available (GA) and not currently scheduled for deprecation or retirement in our tracker.

Does GPT-5.6 Cyber support image or vision input?

Yes — GPT-5.6 Cyber accepts image input. Listed modalities: text, image.

Can I self-host GPT-5.6 Cyber?

No — GPT-5.6 Cyber is proprietary. Its weights are not publicly released, so it is available only through OpenAI's API (or hosted partners), not for self-hosting.

What is the API model string for GPT-5.6 Cyber?

The API model identifier for GPT-5.6 Cyber is "gpt-5.6-cyber" when calling OpenAI's API.

What is GPT-5.6 Cyber's knowledge cutoff?

GPT-5.6 Cyber's training knowledge cutoff is 16 Feb 2026.