GPT-5.6 Sol

GPT-5.6 Sol

Quick answer: GPT-5.6 Sol is OpenAI's flagship frontier model, released July 9, 2026. It achieves state-of-the-art results across coding, cybersecurity, and reasoning benchmarks — including 94.6% on GPQA Diamond, 88.8% on TerminalBench 2.1, 84.5% on CyberGym, and 64.6% on SWE-Bench Pro — while delivering stronger performance per dollar than prior models. Priced at $5 input / $30 output per 1M tokens via the OpenAI API.

At a Glance

Where GPT-5.6 Sol leads

• SOTA on TerminalBench 2.1 at 88.8% — best among evaluated coding agents
• GPQA Diamond 94.6% — among the highest scientific reasoning scores reported
• CyberGym 84.5% — top result for a standard (non-cyber-specialised) model
• SWE-Bench Pro 64.6% — strong long-horizon software engineering performance
• MRCR v2 91.5% (8-needle, 256K–512K range) — competitive long-context retrieval
• Significantly fewer output tokens and lower latency than comparable frontier models

Where it lags

• Highest-tier reasoning (max, ultra) adds latency and token cost
• Closed API model with no open weights
• ultra mode (multi-agent) requires Pro or Enterprise plan access

Best for: Agentic coding, terminal workflows, cybersecurity analysis, and long-horizon professional tasks where intelligence-per-dollar matters.

What GPT-5.6 Sol Is

GPT-5.6 Sol is the flagship of OpenAI's GPT-5.6 model family, which also includes Terra (balanced, everyday work) and Luna (fastest and most cost-efficient). The family introduces a new naming convention where the number identifies the generation and Sol/Terra/Luna are durable capability tiers that can advance independently. GPT-5.6 Sol represents OpenAI's strongest general-purpose model at launch, trained to deliver more useful work per token — a deliberate shift toward efficiency rather than raw scale.

The model introduces max and ultra reasoning effort levels. max gives Sol additional time to reason, explore alternatives, and revise its approach before responding. ultra goes further by coordinating four parallel subagents by default, trading higher token use for stronger results and faster completion on demanding multi-step tasks. This makes GPT-5.6 the first OpenAI model to natively expose a multi-agent orchestration mode at the API level.

GPT-5.6 Sol is also OpenAI's strongest cybersecurity model to date, achieving 84.5% on CyberGym under standard conditions. A Trusted Access for Cyber programme (OpenAI Daybreak) provides verified defenders with more permissive access to advanced cyber capabilities. GPT-5.6 launches with OpenAI's most extensive safety evaluation to date, including over 700,000 A100e GPU hours of automated red-teaming.

Specifications

Pricing

Cache writes are billed at 1.25× the uncached input rate; cache reads receive a 90% discount. Pricing per OpenAI pricing page as of July 2026.

Benchmark Scores

Scores are for GPT-5.6 Sol at max reasoning effort unless otherwise noted. Source: OpenAI GPT-5.6 announcement, July 9, 2026.

Model Family

Source: Artificial Analysis, July 2026.