OpenAI has announced two interlocking moves that together mark one of the sharpest reductions in frontier Large Language Model (LLM) pricing on record: deep cuts to the application programming interface (API) cost of two models in its GPT-5.6 family, and a new program extending free access to those models to 100,000 academic researchers.
Price Reductions
<cite index="1-1">OpenAI has cut the price of GPT-5.6 Luna by 80%, reducing its API cost from $1 to $0.20 per million input tokens and from $6 to $1.20 per million output tokens.</cite> <cite index="6-5">GPT-5.6 Terra's rates fell to $2 per million input tokens and $12 per million output tokens, from $2.50 and $15 — a 20% reduction.</cite> <cite index="6-6">Pricing for Sol, the most capable model in the family, remains unchanged.</cite> Both reductions took effect on July 30, 2026.
<cite index="4-10,4-11">According to OpenAI, the reduction is possible thanks to efficiency improvements resulting from the optimization of the models and of the infrastructure that runs them, with each response now requiring less time, fewer tokens, and fewer resources while offering the same results.</cite>
A notable element of those efficiency gains is AI-assisted self-optimization. <cite index="26-5,26-6,26-7">GPT-5.6 Sol found work that could be precomputed, avoided, or parallelized, and — via Codex — autonomously rewrote and optimized production GPU (Graphics Processing Unit) kernels written in Triton and Gluon, two open-source GPU programming languages maintained by OpenAI.</cite> <cite index="25-1">The kernel work helped reduce the end-to-end cost of serving the model by 20%, while its experiments increased token-generation efficiency by more than 15%.</cite> OpenAI president Greg Brockman noted publicly that "applying GPT-5.6 Sol to optimize our production services is one reason the model offers high performance at a low cost."
<cite index="3-5">GPT-5.6 Luna's combined token price of $1.40 per million tokens now undercuts Google's Gemini 3.5 Flash-Lite at $2.80 and Gemini 3.6 Flash at $9.</cite> <cite index="1-4">The strategic reduction highlights growing competition from open-weight models, compelling proprietary AI providers to lower costs.</cite>
ChatGPT for Academic Researchers
Concurrently, <cite index="13-1">OpenAI is giving 100,000 academic researchers free access to ChatGPT's most advanced AI models to accelerate scientific research, collaboration, and discovery.</cite> The program is called ChatGPT for Academic Researchers.
<cite index="13-2,13-3">OpenAI is starting with 10,000 researchers this summer, with access already available at institutions such as the Institute for Advanced Study (IAS) and École normale supérieure (ENS), and plans to expand to 100,000 researchers through 2027.</cite>
<cite index="9-7">Selected researchers receive GPT-5.6 Sol Pro across ChatGPT, ChatGPT Work, and Codex, plus four collaborator invites and business-grade privacy on their data.</cite> <cite index="16-4">Participants will get a year of access roughly equivalent to OpenAI's $200-per-month ChatGPT Pro account.</cite> <cite index="13-5">Workspaces include business-grade privacy and security protections, and data is not used to train models by default.</cite>
<cite index="16-3">The program is aimed at researchers across scientific and mathematical fields, including biology, chemistry, computer science, engineering, mathematics, and physics.</cite> <cite index="17-5">Researchers will also receive higher usage limits, larger context windows, expanded access to deep research, and more than 75 life science skills covering areas such as genomics, protein modeling, sequencing, and drug discovery.</cite>
<cite index="11-4">OpenAI said the initiative is part of a commitment of more than $250 million through 2027 to support external scientific research and discovery.</cite> <cite index="10-3">That portfolio includes NextGenAI, a $50 million consortium launched in March 2025 with 15 founding research institutions — among them Harvard, MIT, Caltech, Oxford, and Howard University — providing research grants, compute funding, and API access.</cite>
Market Context
<cite index="6-2">OpenAI disclosed that its models now reach more than one billion active users and more than two million businesses.</cite> <cite index="6-10,6-11">The company is operating at a loss as it continues to spend on model development and infrastructure, with financials showing revenue of $13.07 billion in 2025 against a net loss of $38.5 billion.</cite>
<cite index="1-6">The foundation model market is increasingly resembling an infrastructure industry, where scale, capital, and operational efficiency are paramount for dominant players.</cite> <cite index="25-2">OpenAI notes that as its models improve and work more autonomously, the company's ability to improve efficiencies accelerates</cite> — a dynamic that could sustain further price compression across the sector.