OpenAI cuts GPT-5.6 Sol API pricing by up to a third

OpenAI has cut API pricing for GPT-5.6 Sol, its current-generation frontier model, dropping input token costs by 20% and output token costs by 33%. The change, logged in the OpenAI API changelog on August 20, 2026, makes GPT-5.6 Sol meaningfully cheaper to run at scale for developers who have already integrated it into production agents and applications.

What's new

Per OpenAI's own changelog: "GPT-5.6 Sol now costs $4 per million input tokens and $20 per million output tokens, representing 20% lower input pricing and 33% lower output pricing."

• New input price: $4 per million tokens (down from $5)
• New output price: $20 per million tokens (down from roughly $30)
• The reduced pricing is explicitly framed as promotional, with OpenAI committing to hold it through at least November 21, 2026
• The update landed alongside other changelog entries the same week, including a Prompt Caching dashboard and transparent-background support for image generation

OpenAI did not attach the price cut to any model capability change — GPT-5.6 Sol's underlying weights and behavior are unchanged. This is a pure cost adjustment, not a new release.

Context

GPT-5.6 Sol sits in OpenAI's current model lineup alongside GPT-5.6 Luna and other variants that have been rolling out across partner platforms this month, including a general-availability push on Amazon Bedrock's cross-Region inference and adoption by products like Replit's new Free Mode. Token pricing has become an active front of competition among frontier labs this year, with Anthropic, Google, and xAI all adjusting rate cards as inference costs fall and rival open-weight models from DeepSeek, Qwen, and others apply price pressure from below.

OpenAI has periodically used time-boxed promotional pricing — rather than permanent list-price cuts — as a lever to keep usage on newer model variants competitive without permanently committing to a lower margin, and this cut fits that pattern.

Why it matters

For developers and businesses running high-volume workloads on GPT-5.6 Sol, a 20-33% reduction in per-token cost directly lowers the cost of serving agentic and long-context applications, which tend to be output-token-heavy — making the output-side cut of a third the more consequential of the two. Because the discount is promotional and time-limited, teams that build cost assumptions around it into November should watch for whether OpenAI converts it to permanent pricing or lets it lapse. The move also signals that OpenAI is treating token economics as a competitive lever in the current model generation rather than waiting for the next flagship release to reset prices.