OpenAI Cuts GPT-5.6 Sol API Pricing
OpenAI has reduced API pricing for its frontier GPT-5.6 Sol model by more than 20%. The temporary reduction targets developers and takes effect as competition intensifies across the AI industry.
The new pricing applies for the next three months. Moreover, the reduction affects API usage and eligible credits for ChatGPT Work and Codex. However, OpenAI says Pro, Plus and Business subscription pricing remains unchanged.
GPT-5.6 Sol Gets Lower API Pricing
GPT-5.6 Sol now costs $4 per one million input tokens. Its output price has also fallen to $20 per one million tokens. Previously, the corresponding rates were $5 and $30.
Therefore, input pricing has dropped by 20%, while output pricing has fallen by one-third. The combined reduction exceeds 20% when considering the model’s previous pricing structure.
The lower rates apply to standard short-context usage. Developers using longer contexts should check the applicable pricing tier because costs can differ. OpenAI’s API documentation also lists separate rates for cached inputs and longer context windows.
GPT-5.6 Sol remains OpenAI’s flagship model within the GPT-5.6 family. The company positions it for complex professional work, including coding, research, science, cybersecurity and agentic workflows.
The model also supports tool use through the Responses API. Furthermore, developers can use capabilities such as computer use, web search, file search and code execution.
OpenAI Expands Its Price-Performance Push
The Sol reduction follows earlier cuts to OpenAI’s smaller GPT-5.6 models. In July, OpenAI reduced Terra pricing by 20% and Luna pricing by 80%.
OpenAI has also focused on improving model efficiency. The company said production GPU kernel improvements reduced serving costs by 20%, while speculative decoding improved token-generation efficiency by more than 15%.
Consequently, the latest price changes extend a broader effort to improve the economics of advanced AI. For developers, lower inference costs can make sustained agentic workloads more practical.
The timing also reflects a more competitive market. Reuters reported that OpenAI’s latest reduction comes as the company faces growing competition from Anthropic and Chinese AI developers.
Meanwhile, rival pricing remains an important benchmark for enterprise developers. Reuters reported that Anthropic lists its frontier Claude Fable 5 model at $10 per million input tokens and $50 per million output tokens. It also lists Claude Opus 5 at $5 per million input tokens and $25 per million output tokens.
Lower Costs Could Support Agentic Development
The pricing change is particularly relevant to developers building AI agents. These systems can generate large volumes of model calls while reasoning, using tools and completing multi-step tasks.
Therefore, inference economics can directly affect whether an agent remains commercially viable at scale. Lower output costs are especially significant because reasoning-heavy applications can generate substantial output volumes.
GPT-5.6 Sol was designed for long-horizon agentic work. OpenAI says the model can coordinate tools, process intermediate results, and monitor progress as work develops.
The company has also introduced multi-agent capabilities in beta. These allow GPT-5.6 to run concurrent subagents and synthesise their work within a single request.
However, the Sol discount is currently temporary. OpenAI’s stated three-month window means developers must assess whether their applications remain economical after promotional pricing ends.
For now, the reduction gives developers more room to test complex workloads and expand production deployments. More importantly, it signals that frontier-model competition is increasingly being fought on both capability and inference economics.
