OpenAI frees ChatGPT for 1 billion users as DeepSeek hikes prices

The AI industry's two pricing strategies collided this week as OpenAI gave away its current-generation model while DeepSeek told developers to expect price hikes.

OpenAI removed text-chat limits for all ChatGPT users on Aug. 6, making GPT-5.6 Luna the default model for Free and Go tiers as weekly active users crossed 1 billion, while DeepSeek notified API customers of imminent price increases.

"GPT-5.6 Sol delivers more focused answers, adapts its level of detail to the question, and avoids unnecessary formatting," OpenAI said in its announcement, describing the upgraded model now available to Plus and Pro subscribers.

The free tier runs Luna, the smallest of three GPT-5.6 models launched July 9, with unlimited text conversations and a new "Think" button for harder questions arriving next week. Plus and Pro users get a reasoning slider to control how much effort the model applies. OpenAI's internal evaluation found factual errors 62 percent less common with Luna and 68 percent less common with Sol compared with GPT-5.5 Instant.

The move marks a pivot from subscription revenue to platform economics. OpenAI posted $5.7 billion in Q1 2026 revenue with a non-GAAP operating loss margin of -122 percent and expects a $14 billion annual net loss. With 50 million paying subscribers out of 1 billion weekly users, the company is betting that free access will drive advertising and enterprise API revenue.

Free access is a platform play, not charity

OpenAI's economics explain the strategy. Advertising reached $100 million annualized within six weeks of launch, and enterprise API customers now contribute more than 40 percent of revenue. Codex has 5 million weekly users. Inference costs per token have fallen sharply over the past 18 months as architecture optimization, quantization, and faster inference engines compound, making free access economically viable.

Luna is not OpenAI's most capable model — it handles everyday conversation, simple writing, and basic translation, while complex reasoning and multi-step coding still require the Sol tier. But by putting a "good enough" model in front of the largest possible audience, OpenAI is replicating the Google search playbook: free entry, monetize the surrounding services.

The competitive pressure is real. Apptopia's U.S. mobile-app data shows ChatGPT's share of daily active users declining since November 2025, while Anthropic's Claude and Meta AI gained ground. Globally, downloads of the most popular GenAI chatbot apps recorded their first negative-growth month in July 2026.

DeepSeek's servers can't keep up

DeepSeek's situation is the mirror image. V4 Flash topped OpenRouter's weekly global call rankings with 7.22 trillion tokens processed, and on Aug. 1 alone the model handled 8 trillion tokens. Peak-hour latency and timeouts became routine, prompting the company to introduce peak/valley pricing in mid-July that doubled rates during high-demand windows. On Aug. 6, DeepSeek told API users to expect a "significant" price increase.

The problem is compute scarcity, not strategy. Ultra-low pricing attracted massive volumes of low-value traffic that crowds out the enterprise and developer workloads DeepSeek needs to monetize. Doubao, ByteDance's AI assistant with 345 million monthly active users, followed the same logic when it launched paid tiers on June 24 at 68, 200, and 500 yuan per month while keeping basic features free.

The divergence creates a three-layer market structure. Free, commodity-level conversation sits at the bottom — OpenAI's Luna occupies this layer to capture default usage. Professional reasoning — code generation, data analysis, legal and medical assistance — occupies the middle, where DeepSeek and Doubao are pushing paid tiers. Agentic execution, where AI completes tasks rather than answering questions, sits at the top with outcome-based pricing. OpenAI's 1 billion weekly users give it the distribution advantage, but the -122 percent operating margin shows the cost of that ambition. DeepSeek's price hike tests whether Chinese AI firms can convert usage into sustainable revenue before compute constraints force the issue.

This article is for informational purposes only and does not constitute investment advice.