OpenAI Slashes GPT-5.6 API Prices: Luna Down 80%, Terra Down 20% (New Pricing Breakdown)

OpenAI cut API prices on two of its three GPT-5.6 models on July 30, 2026. GPT-5.6 Luna now costs 80% less: $0.20 per million input tokens and $1.20 per million output tokens. GPT-5.6 Terra dropped 20% to $2 input and $12 output. GPT-5.6 Sol, the flagship reasoning model, keeps its $5 and $30 rates but gains a new Fast mode that runs 2.5x faster at double the standard price.

Try Apidog today

If you built your cost model around the launch-day GPT-5.6 pricing, update it now. This guide covers the new rate card, the impact on API routing, and a practical way to compare Sol, Terra, and Luna using your own production-like prompts.

The fastest way to validate a model-routing change is to send the same request through each tier and compare quality, latency, and token usage. Tools such as Apidog let you run those requests side by side before changing production defaults.

The new GPT-5.6 rate card

Here is the before-and-after pricing per million tokens:

Model	Old input	New input	Old output	New output	Change
GPT-5.6 Sol	$5.00	$5.00	$30.00	$30.00	No change
GPT-5.6 Terra	$2.50	$2.00	$15.00	$12.00	-20%
GPT-5.6 Luna	$1.00	$0.20	$6.00	$1.20	-80%

The new prices took effect on July 30, 2026 and apply to the standard API tier. The spread between the top and bottom of the family grew from 5x at general availability on July 9 to 25x.

That changes routing decisions:

• Choosing Luna over Sol previously saved 80%.
• Choosing Luna over Sol now saves 96% on list price.
• Terra remains a middle tier, but its 20% reduction may change your quality-per-dollar threshold.

Prompt-caching discounts still apply on top of these rates. OpenAI has not published separate cached-input prices for the new Luna and Terra rates in the announcement, so check the official pricing page before updating forecasts that include cache reads.

What changed for Sol

Sol's standard list price is unchanged. Instead, OpenAI introduced Fast mode:

• Speed: 2.5x faster token generation
• Price: 2x the standard Sol rate
• Migration: Fast mode replaces the previous Priority Processing tier

If you currently use Priority Processing, review your production configuration and invoice mapping. Your integration may need a configuration update even though the base Sol model price did not change.

The practical routing rule is straightforward:

• Use Sol when you need flagship reasoning quality.
• Use Sol Fast mode when latency matters enough to justify the premium.
• Route suitable workloads to Terra or Luna when throughput and cost matter more.

Why OpenAI cut prices

OpenAI describes the cuts as efficiency gains passed through to API customers. The announcement cites four contributors:

• Improved hardware routing across its inference fleet
• Better production inference software
• Smarter context-caching algorithms
• Production serving kernels rewritten by GPT-5.6 Sol

OpenAI says Sol autonomously rewrote serving kernels, reducing end-to-end serving overhead by 20% and improving token-generation efficiency by more than 15%.

There is also a competitive factor. VentureBeat’s coverage frames the cut as a response to Google’s Flash tier.

Luna’s combined list price is now $1.40 per million tokens:

• Gemini 3.5 Flash-Lite: $2.80 combined
• Gemini 3.6 Flash: $9 combined
• GPT-5.6 Luna: $1.40 combined

List price is only one input to a routing decision. Quality, latency, output length, and retry rates can change the real cost of a workload.

What the cut means for your workloads

The 80% Luna reduction is the change most likely to affect architecture decisions.

Agent loops get cheaper

Agentic workloads consume tokens through repeated planning, tool calls, tool-result processing, and retries. At Luna’s previous $1 input and $6 output pricing, these loops could still be expensive at scale. At $0.20 input and $1.20 output, Luna becomes a more viable default for cost-sensitive agent paths.

Use this as a trigger to re-evaluate:

• Tool-calling agents
• Multi-step extraction pipelines
• Automated research and summarization flows
• Retry-heavy workflows

Move classification, extraction, and summarization down-tier

If you previously kept simple tasks on Terra because Luna’s quality margin seemed too narrow, run the evaluation again. The model quality did not change, but the cost profile did.

For example, a workload costing $100 per day on Terra would cost about $11 per day on Luna at the new rates, assuming comparable token usage.

Terra remains a practical default

The Sol vs Terra vs Luna breakdown recommended Terra as a sensible default for everyday work. That recommendation still holds, and Terra now costs 20% less.

Reasoning effort still drives output cost

GPT-5.6 exposes six reasoning effort levels. Higher effort can generate more output tokens, and output tokens remain the more expensive direction on every tier.

Do not optimize only the model ID. Also evaluate reasoning effort:

1. Start with the lowest effort that meets your quality target.
2. Record output-token usage for each effort level.
3. Promote effort only for requests that fail quality checks.

How to re-check your own numbers

Do not rely only on vendor percentages. Measure your actual traffic.

1. Sample representative requests

Build a dataset that includes:

• Common user prompts
• Long-context requests
• Tool-calling flows
• Failure-prone or retry-heavy requests
• Quality-sensitive edge cases

Avoid evaluating only short, clean prompts. They rarely reflect production token usage.

2. Run every request against all three tiers

Send the same request to:

gpt-5.6-luna
gpt-5.6-terra
gpt-5.6-sol

Capture at least:

• Input tokens
• Output tokens
• Total latency
• Response quality score
• Retry or failure status

3. Calculate request-level cost

Use the model’s input and output rates independently:

request_cost =
  (input_tokens / 1_000_000 * input_rate) +
  (output_tokens / 1_000_000 * output_rate)

For Luna:

request_cost =
  (input_tokens / 1_000_000 * 0.20) +
  (output_tokens / 1_000_000 * 1.20)

For Terra:

request_cost =
  (input_tokens / 1_000_000 * 2.00) +
  (output_tokens / 1_000_000 * 12.00)

For Sol:

request_cost =
  (input_tokens / 1_000_000 * 5.00) +
  (output_tokens / 1_000_000 * 30.00)

4. Compare quality and cost together

Create a simple evaluation table:

Model	Quality score	Input tokens	Output tokens	Estimated cost	Latency
Luna
Terra
Sol

Then define routing rules based on the results. For example:

Use Luna by default.
Escalate to Terra when the Luna quality score is below the threshold.
Escalate to Sol only for high-value or reasoning-intensive requests.
Use Sol Fast mode only where latency requirements justify its premium.

Apidog can speed up the comparison step: define the chat completion endpoint once, store the model ID in an environment variable, and run the same scenario against all three tiers. The response viewer exposes the usage block for each call, giving you actual input and output token counts.

If you are integrating GPT-5.6 for the first time, see this guide to using the GPT-5.6 API for authentication, model IDs, and request shape. You can also mock the endpoint while waiting for budget approval, then switch to live traffic without rewriting tests.

Download Apidog free to run the comparison.

The bigger picture: the price war is on

This is the third major price move at the low-cost end of the frontier this summer. Google refreshed its Flash tier in July, Anthropic launched Claude Opus 5 at half of Fable 5’s price, and OpenAI cut Luna below both Gemini Flash options.

Flagship models still differ in quality, but workhorse-tier routing increasingly favors the cheapest model that meets your evaluation threshold.

For API teams, model selection should not be an annual architecture decision. Treat it as operational infrastructure:

• Keep the model ID in configuration, not application logic.
• Maintain a standing evaluation suite.
• Record token usage and latency by model.
• Re-run comparisons when vendors change pricing or model behavior.
• Use routing thresholds that can be changed without redeploying your application.

FAQ

What are the new GPT-5.6 API prices?

As of July 30, 2026:

• Luna: $0.20 per million input tokens and $1.20 per million output tokens
• Terra: $2 per million input tokens and $12 per million output tokens
• Sol: $5 per million input tokens and $30 per million output tokens

Sol also gained a Fast mode that costs double the standard rate for 2.5x faster processing.

Did GPT-5.6 Sol get a price cut?

No. Sol’s standard rates are unchanged.

The Sol change is structural: Fast mode replaces Priority Processing, trading 2x the price for 2.5x speed.

Is GPT-5.6 Luna now cheaper than Gemini Flash?

On list price, yes. Luna’s combined input-plus-output rate is $1.40 per million tokens, compared with $2.80 for Gemini 3.5 Flash-Lite and $9 for Gemini 3.6 Flash.

Price is not quality. Run both models against your own evaluation set before switching. For a free starting point, see how to use GPT-5.6 for free.

Do the price cuts apply automatically?

Yes. The new rates apply to standard API usage from July 30, 2026 without code changes.

You only need to take action if you use Priority Processing on Sol, which is being replaced by Fast mode.

What to do this week

1. Update your cost model using the new Luna and Terra rates.
2. Re-evaluate workloads that can move to Luna at 96% below Sol’s list price.
3. Check whether Terra’s 20% reduction changes your routing thresholds.
4. Review Sol Priority Processing usage and plan the Fast mode migration.
5. Run real requests through Luna, Terra, and Sol.
6. Compare usage data, latency, and quality before changing defaults.

Use one test scenario in Apidog, switch among three model IDs, and use the returned usage values to determine what the price cut is worth for your traffic.