OpenAI has cut GPT-5.6 Sol’s API and credit pricing by over 20%, dropping the standard rate to $4 per million input tokens and $20 per million output tokens for a three-month promotional period starting August 22. The reduction is Sol’s first official price cut since its July launch and the second across the GPT-5.6 family in under a month.
What Changed
Sol’s standard rate falls from $5 to $4 for input tokens and from $30 to $20 for output tokens – a 20% cut on input and a 33% reduction on output. Cached input drops from $0.50 to $0.40 per million tokens. Batch processing falls to $2 input and $10 output, while Fast mode sits at $8 input and $40 output – all reflecting the same proportional reductions.
The cuts apply to the API, Codex credits, ChatGPT Work plans, and Amazon Bedrock. Pro, Plus, and Business subscription usage remains unchanged. OpenAI says the promotional rate is valid at least through November 21, 2026.
This follows OpenAI’s July 30 price cuts for the smaller models in the family: GPT-5.6 Luna dropped 80% and Terra fell 20%. OpenAI attributed those earlier reductions to self-optimization work where GPT-5.6 improved its own runtime efficiency, yielding 20% lower serving costs through production GPU kernel improvements and 15% better token-generation efficiency via improved speculative decoding.
Context and Competition
The Sol cut lands in an increasingly competitive pricing landscape. DeepSeek raised V4 API prices 50 to 1,100 percent in mid-August, moving in the opposite direction. Chinese models from Alibaba and Moonshot AI continue to undercut Western competitors on raw token pricing. Meanwhile, Anthropic’s Claude Sonnet 5 pricing has held steady through the same window.
Developer reaction has been mixed. While cheaper API access is broadly welcomed, some developers on OpenAI’s community forum pointed out that rate limits – not pricing – remain the binding constraint. A 20% price cut doesn’t help if you’re already hitting a 5-hour rolling window cap before you hit a spending ceiling.
The three-month window also signals OpenAI expects a new model release before November, which would likely supersede the promotional pricing. Analysts note this is the first time OpenAI has directly repriced a model’s own API rate since 2022, rather than adding a faster tier alongside an unchanged base price.
For teams building on Sol, the reduction represents genuine cost savings on the current model. But the broader picture – aggressive pricing cuts alongside strict rate limits and an experimental release cadence – suggests OpenAI is prioritizing developer acquisition while it sorts out capacity constraints. The real question is whether the promotional pricing becomes permanent or gets quietly replaced when the next model ships.
discussion