OpenAI just made a massive move, slashing the prices for its latest GPT-5.6 models by up to 80%. Interestingly, this huge discount comes just three weeks after the model’s initial launch. Furthermore, OpenAI followed up this aggressive pricing strategy on July 31 by revealing it has officially surpassed 1 billion active users.
OpenAI Slashes GPT-5.6 Prices Amid Chinese Pressure
The price war in artificial intelligence is heating up quickly. Cheaper Chinese open-weight models are putting heavy pressure on industry giants like OpenAI and Anthropic to justify their costs. Consequently, OpenAI took aggressive action this week.
The company cut the price of GPT-5.6 Luna, its fastest and lowest-cost model, by a staggering 80%. Luna now costs just $0.20 per million input tokens and $1.20 per million output tokens. Meanwhile, the mid-range GPT-5.6 Terra received a 20% price reduction. Terra will now cost $2.00 for input tokens and $12.00 for output tokens.
However, the flagship GPT-5.6 Sol model keeps its standard price. Instead of a discount, OpenAI introduced a new “Fast mode” for Sol. This new mode delivers up to 2.5 times the standard processing speed at twice the standard price, without sacrificing any underlying intelligence.
AI Optimizing AI for Better Efficiency
How did OpenAI achieve these massive discounts so soon after launch? Surprisingly, the company did not alter the core intelligence models at all. Instead, OpenAI heavily optimized the surrounding software systems.
In fact, the company used GPT-5.6 Sol to optimize its own production software. This clever move reduced end-to-end serving costs by 20%. Additionally, the engineering team improved speculative decoding, which boosted token-generation efficiency by over 15%.
These system upgrades also yielded massive benchmark improvements. By upgrading retained reasoning and context management, OpenAI pushed GPT-5.6 Sol’s score on the ARC-AGI-3 task set from 13.3% to 38.3%. Crucially, the system achieved this dramatically higher score while consuming six times fewer output tokens.
A New Standard: Total Task Cost
Beyond pricing, OpenAI shared massive internal growth metrics. The company now reports over 1 billion active users and more than 2 million business clients. However, they intentionally did not specify whether this user count tracks weekly or monthly activity.
Still, overall user engagement is skyrocketing. Six months after signing up, users send 50% more daily messages and use the platform for twice as many types of work. Moreover, the nature of that work is shifting. ChatGPT Work is moving away from simple questions and toward complex, multistep tasks. Internally, OpenAI relies heavily on this agentic approach. Right now, agentic workflows through Codex drive 99.8% of the company’s weekly output tokens, with their Finance team leading the charge.
Ultimately, OpenAI wants the tech industry to stop focusing solely on raw token prices. Instead, they urge developers to measure the “cost of a successful outcome”. A slightly more expensive model that gets a complex task right immediately actually costs less than a cheap model requiring endless retries, tool calls, and human review. Moving forward, OpenAI plans to expand its infrastructure strictly based on this credible demand and proven efficiency, rather than blindly building data centers.
