Skip to main content
  • Published: July 30, 2026
  • Last verified: August 1, 2026
  • Current status: Effective; new rates apply to gpt-5.6-luna and gpt-5.6-terra, while gpt-5.6-sol pricing is unchanged
Starting July 30, 2026, OpenAI reduced GPT-5.6 Luna API prices by 80% and GPT-5.6 Terra API prices by 20%. LaoZhang API has applied the new rates automatically. Existing users do not need to change API keys, base URLs, model names, request parameters, or integration code.

​
Current prices

All prices below are in US dollars per 1 million tokens for standard requests with no more than 272K input prompt tokens.
This table covers standard GPT-5.6 token rates. Special token groups, service tiers, or other billing modes remain subject to the console model and pricing page and call logs.

​
How are prompts over 272K input tokens priced?

When a request contains more than 272K input prompt tokens, OpenAI applies 2× input pricing and 1.5× output pricing to the entire request. Using the new base rates: The long-context multiplier applies to the full request, not only to tokens above 272K. Use the complete input-token count when estimating cost, then verify actual charges in the console call log.

​
What made the price cut possible?

OpenAI attributes the lower prices to efficiency improvements across the models, inference systems, and agentic harness, including production software, hardware routing, and context management. In a human-led engineering process, GPT-5.6 Sol also helped rewrite and optimize production kernels and run token-generation experiments. OpenAI reports that the kernel work reduced end-to-end serving cost by about 20%, while the experiments improved token-generation efficiency by more than 15%.

​
Who is affected?

  • gpt-5.6-luna users receive the new price automatically and can reevaluate the cost of bulk classification, extraction, and high-throughput agent workloads.
  • gpt-5.6-terra users receive the 20% reduction automatically.
  • Standard and long-context prices for gpt-5.6-sol remain unchanged.
  • API keys, the https://api2.laozhang.ai/v1 base URL, model IDs, and Chat Completions or Responses API request formats do not change.

​
What do users need to do?

  1. No code changes are required: Keep using the existing API key, endpoint, and model name.
  2. Update cost forecasts: Recalculate Luna and Terra batch jobs and agent workflows with the new rates.
  3. Account for long context: If an input prompt can exceed 272K tokens, estimate the full request at the long-context rate.
  4. Verify actual billing: Confirm the token group on the console model and pricing page, then check token usage and charges in call logs.

​
Frequently asked questions

​
Do I need a new API key, base URL, or model name?

No. This update changes pricing only. Existing LaoZhang API keys, base URLs, model IDs, and request formats continue to work.

​
Is only the portion above 272K tokens charged at the higher rate?

No. Once the input prompt exceeds 272K tokens, the 2× input and 1.5× output multipliers apply to the entire request.

​
Did gpt-5.6-sol become cheaper?

No. gpt-5.6-sol remains $5.00 input, $0.50 cached input, and $30.00 output per 1 million tokens. For prompts over 272K input tokens, input and output are $10.00 and $45.00 respectively.

​
Are earlier calls recalculated at the new price?

This announcement documents the current rates effective July 30, 2026. It does not retroactively change recorded charges for earlier calls; use existing console billing records for historical usage.

​
How can I confirm that my token uses the new price?

Open the console model and pricing page, confirm the model and token group, then send one low-volume request and review its input tokens, output tokens, and actual charge in the call log.