On June 25-26, OpenAI launched a limited preview of its new frontier model series "GPT-5.6." Citing security concerns in cyber and bio domains, a US government request has confined the initial rollout to government-approved partners under a phased deployment.
June 26 · OpenAI · Limited Preview
GPT-5.6 lands first with 20 partners — and the U.S. government
OpenAI opens a staggered preview of its three-tier "GPT-5.6" family — flagship Sol , balanced Terra , fast-and-cheap Luna — gated behind prior government security review, with a broader release targeted for next week.
~6 wks
Release cadence — GPT-5.6 arrives ~2 months after GPT-5.5
~20
Approved partners in the first limited preview
1.5M
Context window (leaked, unconfirmed) — +43% vs 5.5
Three tiers · API price per 1M tokens
Output price drawn to scale (taller column = costlier)
$15
Terra
Balanced · in $2.50
OpenAI claims Terra is effectively half the price of GPT-5.5 — while developers warn flagship costs may run up to double 5.5.
Staggered rollout, gated by review
U.S. gov review
Pre-release security check
→
~20 partners
Limited preview now
→
Broad release
ChatGPT & Codex next
Driven by concern over frontier models' cyber capabilities. The model has not hit the "Cyber Critical" threshold — it can spot bugs in Chromium & Firefox but is restricted from autonomously building end-to-end exploits.
Developer hopes
1.5M context for long docs & large codebases
Stronger agentic coding workflows
10–15% better token efficiency vs 5.5
Pricing edge vs rivals like Fable 5
Concerns
Flagship cost may reach ~2× GPT-5.5
No broad access yet — gated by review
Context window unconfirmed officially
Benchmarks & real-world tests still pending
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…