OpenAI Officially Launches GPT-5.6
OpenAI has officially launched GPT-5.6, which is divided into three tiers: Sol as the flagship model, focusing on complex reasoning and long-term autonomous work; Terra as the cost-effective option, with performance close to the previous flagship GPT-5.5 but at half the price; and Luna as the lightweight version, which is fast and low-cost.
In terms of API pricing, Sol charges $5 per million tokens for input and $30 for output, while Luna charges $1 for input and $6 for output. ChatGPT free and Go users will default to using Terra, while Plus and higher plans can choose their model.
Sol introduces a new max deep thinking mode and an ultra multi-agent parallel mode, the latter currently limited to Pro and Enterprise users.
Source: Public Information
ABAB AI Insight
OpenAI has previously established a multi-tier strategy through models like GPT-4o, and this launch of GPT-5.6 with three tiered options continues its path of refined pricing and coverage of different user needs. The ultra multi-agent mode of Sol represents an important step towards autonomous workflows.
From a capital perspective, OpenAI attracts mass users with the low-priced Terra, while serving high-end demands with the premium-priced Sol, rapidly expanding revenue scale. The motivation is to balance costs and revenues while accelerating ecosystem expansion, specifically through differentiated API pricing that locks in enterprise and developer budgets, while the ultra mode drives conversion to high-paying plans.
Similar to competitors like Anthropic and Google with their multi-model strategies, OpenAI is currently in a mature stage of model commercialization, with tiered pricing and new reasoning modes reinforcing its market dominance.
Essentially, this represents a transfer of pricing power: the multi-tier architecture of GPT-5.6 is reshaping the pricing structure of the AI model market, with a mechanism that finely matches performance and price to cover the entire user spectrum, driving capital towards platforms that can simultaneously control high-end reasoning and mass traffic, further consolidating OpenAI's industry control.
ABAB News · Cognitive Laws
Flagship expensive, cost-effective moderate, and civilian cheap is the optimal capital structure to cover the entire market.
Multi-agent parallel thinking is closer to real work scenarios than a single model.
When models are tiered by scenario and budget, AI truly transforms from a toy into infrastructure.