OpenAI challenges Chinese models by reducing prices for GPT-5.6 Luna by 80% and Terra by 20%
OpenAI reduces prices for its GPT-5.6 Luna models by 80% and Terra by 20%, questioning the competitiveness of Chinese developments and strengthening its position in the AI market.
Краткий пересказ от QRazy ИИ
- OpenAI has lowered the prices of the GPT-5.6 Luna and Terra models, increasing their accessibility.
- Optimization of infrastructure and algorithms has led to reduced operating costs and increased efficiency.
- Competition with Chinese developers has prompted American companies to reconsider their pricing strategies.
OpenAI has reduced the prices of two models from the GPT-5.6 family. The price of GPT-5.6 Luna has decreased by 80%, while GPT-5.6 Terra has dropped by 20%.
The GPT-5.6 series includes three models. GPT-5.6 Sol is designed for complex reasoning and agent tasks, Terra occupies an intermediate position, and Luna is the most compact and economical model in the family.
Initially, Luna cost $1 per million input tokens and $6 per million output tokens. Now, the prices are $0.20 and $1.20, respectively. For Terra, the cost has decreased from $2.50 to $2 per million input tokens and from $15 to $12 per million output tokens.
The new pricing will also be taken into account when calculating usage volume under the paid subscriptions for Codex and ChatGPT Work.
According to OpenAI, the price reductions were made possible by optimizing the infrastructure after several weeks of the models being used by clients around the world. Improvements to GPU cores have reduced operational costs by 20%, and enhanced speculative decoding has increased token generation efficiency by 15%.
The price of the flagship GPT-5.6 Sol has not changed: the model still costs $5 per million input tokens and $30 per million output tokens. In its API, it will introduce a Fast mode, which will replace Priority Processing. This mode can operate up to 2.5 times faster than the standard mode but will cost twice as much.
OpenAI has also transitioned the Auto-review feature in the ChatGPT and Codex CLI applications from GPT-5.4 to GPT-5.6 Luna. The company expects that this will make the review process about ten times cheaper.
Initially, OpenAI provided access to GPT-5.6 only to a limited number of companies. This occurred in June amid restrictions from the U.S. government. Earlier this month, the models became available to clients worldwide.
Prices are decreasing amid competition with Chinese developers. The Kimi models from Moonshot AI and the GLM family from Z.ai offer tools for reasoning, programming, and agent tasks at significantly lower prices than many leading American models. This forces U.S. companies to reduce inference costs and reassess pricing for developers.