OpenAI GPT-5.6 — официальный релиз
Sol / Terra / Luna — deep dive (2026)

26 июня 2026 — OpenAI ship'ит GPT-5.6 в трёх SKU: flagship Sol, balanced Terra, lightweight Luna (первый solar-system naming scheme). Tech deep dive для AI devs и tech leads: pricing matrix, Max/Ultra inference modes, TerminalBench 2.1 — 91,9% (#1 globally), CTF hit rate 96,7%, Cerebras 750 tok/s с июля, US government lock на preview, head-to-head vs Claude Mythos 5, 6-step access playbook и FAQ. Сейчас доступ только у ~20 pre-cleared partners; public rollout — через несколько недель.

01

Pain point: почему GPT-5.6 недоступен большинству devs

Июнь 2026 — super release month на бумаге. На практике три blocker'а держат модель за paywall/government gate:

  1. 01

    Access gated: по требованию US gov GPT-5.6 доступен только ~20 trusted partners через API/Codex — ни ChatGPT, ни public API для regular users

  2. 02

    Competitive vacuum: Claude Mythos 5 offline с 12 июня (export control), Gemini 3.5 Pro slipped to July — coding agent market без clear SOTA

  3. 03

    Policy entropy: executive order от 2 июня 2026 = precedent gov intervention в model releases; timeline prediction стал harder

Pricing & positioning — raw numbers

ModelTierInputOutputKey metric
GPT-5.6 SolFlagship$5 / M tok$30 / M tokTerminalBench 2.1: 91,9%
GPT-5.6 TerraWorkhorse$2.50 / M tok$15 / M tok~GPT-5.5 perf, −50% cost
GPT-5.6 LunaLightweight$1 / M tok$6 / M tok80% price edge vs Sol tier
warning

Current state: preview-only для ~20 partners. Polymarket odds на full release до 31 июля 2026: 87%.

02

Release context & model stack: Sol / Terra / Luna

Ночью 27 июня 2026 (CST) OpenAI анонсировала GPT-5.6 с solar naming: Sol (Sun), Terra (Earth), Luna (Moon) — flagship / mid / edge tier.

Release прошёл под government constraint: впервые US gov потребовал limited release до wide rollout. CEO Sam Altman compliant, но публично pushback:

Government approval gate не должен стать industry default — он отрезает лучшие tools от users, devs, enterprises и global partners, которым они нужны больше всего.

GPT-5.6 Sol — flagship SKU

Strongest OpenAI model to date. Target workloads: hardcore coding, long-chain cyber research, multi-step agentic pipelines.

Два новых inference mode:

  • Max mode: больше reasoning budget, accuracy ↑, latency ↓ (trade-off осознанный)
  • Ultra mode: multi-agent orchestration — task decomposition → parallel sub-agents → merge output; driver 91,9% на TerminalBench

Pricing: $5/M input, $30/M output (parity с GPT-5.5).

GPT-5.6 Terra — enterprise workhorse

Mass-deploy tier: support bots, internal tools, doc analysis. ~GPT-5.5 quality при 50% lower bill. $2.50/M in, $15/M out.

GPT-5.6 Luna — lightweight edge

Tuned для high-QPS, low-latency: summarization, drafting, cron automation. Первый non-flagship с simultaneous High rating в cyber + bio. $1/M in, $6/M out.

ModelPrimary workloadContextCyber rating
SolComplex coding, security R&D, long agents~1.5M tokHigh
TerraEnterprise docs, support, mass API~1.5M tokHigh
LunaSummary, drafting, automation~1.5M tokHigh
03

Benchmark dump: coding, agents, cyber

TerminalBench 2.1 — code agent SOTA

89 hard CLI planning tasks; measures multi-step tool use, iterative fix loops, task coordination — closest proxy к real agent behavior.

ModelScoreMode
GPT-5.6 Sol91,9% — global #1Ultra (multi-agent)
GPT-5.6 Sol88,8%Standard
Claude Mythos 588,0%Standard
GPT-5.583,4%Standard
Gemini 3.1 Pro Preview70,7%Standard

Sol dethroned Mythos 5 за 17 дней (Mythos topped chart 9 июня). Pre-release context: GPT-5.6 leak roundup.

Agent's Last Exam — long-horizon tasks

ModelTask completion (code mode)
GPT-5.6 Sol50,9% — единственный >50%
GPT-5.6 LunaSlightly above GPT-5.5

Cyber: CTF & ExploitBench

Первая OpenAI product line, где все три SKU trigger High cyber risk tier.

ModelCTF hit rate
Sol96,7%
Terra91,84%
Luna85,19%

ExploitBench: Sol ≈ Anthropic Mythos Preview, но ~⅓ output tokens — materially cheaper security research stack.

shield

Safety boundary: Sol находит vulns/primitives в Chromium/Firefox codebases, но не собирает full weaponized exploit chains autonomously — below OpenAI «Cyber Critical» threshold.

Life sciences: GeneBench v1 & HealthBench

  • GeneBench v1: Sol matches/exceeds GPT-5.5 при меньшем token burn
  • HealthBench Professional: 60.5 pts — +8.7 vs GPT-5.5
04

Cerebras 750 tok/s + government drama

Speed tier: Cerebras deploy с July

GPT-5.6 Sol на Cerebras WSE: до 750 tok/s generation. Baseline flagship: 50–150 tok/s. Math: 5–15× throughput — game changer для realtime coding assistants и streaming UX.

Executive order (2 June 2026)

US gov получает до 30 days pre-release access для security review. Non-binding on paper, effective in practice. 26 June: OpenAI cap at ~20 trusted partners (OSTP + ONCD coordination).

Big Three — release status matrix

LabModelStatus
OpenAIGPT-5.6 Sol/Terra/LunaPreview ~20 partners
AnthropicClaude Fable 5 / Mythos 512 June: export kill switch, global offline
GoogleGemini 3.5 ProSlipped to July (was June)

Sol vs Mythos 5 — spec sheet compare

DimensionGPT-5.6 SolClaude Mythos 5
TerminalBench 2.191,9% (Ultra) / 88,8%88,0%
ExploitBench≈ Mythos Preview, ~⅓ tokensNo public data
Input price$5 / M$10 / M (offline)
AvailabilityLimited preview, GA in weeksExport control — offline
Context window~1.5M tok200K tok

Sol wins coding/cyber benchmarks при 2× cheaper input. Fable 5 still competitive на SWE-bench Pro — full compare после System Card drop. Background: Claude Fable 5 export control breakdown.

05

Access playbook: 6 steps + use-case routing

Timeline: June now vs July expected

  • Now: ~20 trusted partners via API/Codex; ChatGPT locked for public
  • July forecast: ChatGPT GA (Plus/Pro first), public API, Cerebras Sol for enterprise (750 tok/s)

6-step dev checklist

  1. 01

    Watch OpenAI status page: alert на public API unlock

  2. 02

    Hold production baseline: GPT-5.5 или Claude Opus 4.8 до GA

  3. 03

    Pre-map model routing: Sol → agents, Terra → mass API, Luna → lightweight

  4. 04

    Priority eval post-GA: TerminalBench-style pipelines, CTF research, long-context doc RAG

  5. 05

    Model token cost: Ultra только для genuinely hard tasks — burn rate spikes

  6. 06

    Cerebras ROI calc: post-July enterprise channel для 750 tok/s eval

Use-case → model routing table

RequirementPick
Complex codegen, debug, multi-agent tasksSol
Enterprise doc analysis, support, mass APITerra
High-QPS summary, drafting, automationLuna
GPT-5.5-tier on tight budgetTerra (−50% cost)
Latency-critical realtime (post-July)Sol on Cerebras

Reference specs (EEAT)

  • TerminalBench 2.1: Sol Ultra 91,9%, standard 88,8% — Mythos 5: 88,0%
  • CTF: Sol 96,7% / Terra 91,84% / Luna 85,19%
  • Cerebras: 750 tok/s (July), 5–15× vs flagship baseline
  • Red team spend: 700K A100-equivalent GPU-hours automated testing

Pure cloud API = fast model swap, но policy whiplash, long-context bill shock и unpredictable Ultra token burn. Self-host = A100/H100 capex + ops headcount. Для 7×24 AI agents, multi-agent coding pipelines или iOS CI/CD automation в production — NodeMini Mac Mini M4 cloud rental даёт unified memory + Apple Silicon efficiency как stable execution layer. Pricing: тарифы аренды.

FAQ

Частые вопросы

Для public users — нет. ~20 trusted partners через API/Codex. ChatGPT rollout ожидается July 2026; Polymarket: 87% до 31 July.

TerminalBench: 91,9% (Ultra) vs 88,0%. ExploitBench parity при ~⅓ tokens. Mythos 5 силён на SWE-bench Pro. См. export control analysis.

Multi-agent orchestration: decompose → parallel sub-agents → merge. Driver 91,9% TerminalBench; token burn выше — reserve для hard workloads only.

Executive order 2 June 2026 → OSTP/ONCD security review gate. OpenAI capped at ~20 partners; CEO против long-term normalization этого pattern.

С July 2026: до 750 tok/s на GPT-5.6 Sol — 5–15× vs типичные 50–150 tok/s. Initial rollout: selected enterprise only.

Sol: complex coding + multi-step agents. Terra: enterprise docs + mass API. Luna: summary + automation. Execution layer: help center или coding assistants shootout.