OpenAI model family / GPT-5.6 Generally available · July 9, 2026
Sol Terra Luna

GPT-5.6 has three tiers, not one default answer.

OpenAI now prices Standard short-context Terra at $2 input and $12 output per million tokens, and Luna at $0.20 input and $1.20 output. Sol adds Fast mode for premium latency. Flowith availability remains a separate live check.

API price record · July 30, 2026

Terra and Luna cost less

These are OpenAI's Standard API rates per 1 million tokens for short-context requests. Prompts above 272K input tokens, other processing modes, cache writes, tools, data residency, and third-party platforms can use different rates.

Verify live pricing

Terra

gpt-5.6-terra

20% lower from July 30
Input
$2.00
Cached
$0.20
Output
$12.00

Luna

gpt-5.6-luna

80% lower from July 30
Input
$0.20
Cached
$0.02
Output
$1.20

API token prices are separate from ChatGPT Work and Codex subscriptions. OpenAI says subscription prices and quota budgets stayed unchanged, while Terra and Luna now consume fewer credits on paid subscriptions.

Family map

Choose by operating role

The tier names are durable roles. Exact limits, pricing, and features should always be checked in the live OpenAI docs.

S Flagship capability

Sol

gpt-5.6-sol

Complex production work where quality, coding, knowledge work, and difficult multi-step execution matter most.

T Balanced everyday work

Terra

gpt-5.6-terra

Strong general performance when a workflow needs a lower-cost role than the flagship tier.

L Fast, efficient throughput

Luna

gpt-5.6-luna

High-volume or latency-sensitive work where efficiency is the primary model-selection constraint.

Sol processing choice

Fast mode replaces Priority Processing

OpenAI says GPT-5.6 Sol can run up to 2.5× faster in Fast mode than Standard processing at twice the token price, with no intelligence change. It is a processing choice, not a different model.

Standard

Default choice when cost matters more than premium latency.

No Fast mode opt-in required.

Use as the evaluation baseline for latency and completed-task cost.

Fast

High-value, user-facing Sol work where latency is worth a premium.

Set service_tier to "fast"; "priority" remains compatible.

Up to 2.5× faster for Sol at twice the Standard token price; ramp limits can downgrade traffic.

Requests using the legacy priority value continue to work. Fast traffic can be downgraded to Standard when ramp-rate limits apply; inspect the returned service tier and usage data when latency or billing matters.

Availability boundary

Provider access is not Flowith access

OpenAI's announcement establishes availability on its own surfaces. It does not establish a Flowith launch, account entitlement, free-credit offer, or model mapping.

ChatGPT Work

OpenAI says paid-plan Terra and Luna usage now consumes fewer credits; subscription price and quota budgets did not change.

Not evidence of Flowith access.

Codex

Terra remains available to Free and Go; paid plans can choose Terra and Luna, with lower credit consumption for both tiers.

Separate OpenAI product surface.

OpenAI API

OpenAI documents Sol, Terra, and Luna model IDs for API use.

Does not confirm a Flowith model mapping.

Flowith

No internal launch evidence was found for GPT-5.6 during this route review.

Verify in the live workspace model selector.

Selection protocol

Three checks before you switch

01

Choose the tier by role

Start with Sol for frontier capability, Terra for balanced work, or Luna for efficient throughput. Do not collapse a multi-model workflow into one tier without evaluation.

02

Confirm the product surface

ChatGPT, Codex, the OpenAI API, and Flowith have separate access and entitlement boundaries. Verify the surface you will actually use.

03

Choose the processing mode

Use Standard as the baseline. Test Fast mode for latency-sensitive Sol requests, then verify the response service tier and billed rate instead of assuming every request received premium processing.

04

Test the real contract

Measure task success, output completeness, latency, token use, tools, and structured-output behavior on representative work before switching.

Primary sources

Read the live OpenAI record

Use the announcement for the availability event and the model guide for current implementation details.

GPT-5.6 launch announcement OpenAI's July 9, 2026 release, family tiers, product availability, and launch-level capability summary. GPT-5.6 model guidance Current API model identifiers, workload roles, reasoning settings, feature notes, and migration guidance. July 30 price update Official Terra and Luna price changes, paid-product credit implications, and the Sol Fast mode announcement. OpenAI API pricing Live Standard, Batch, Flex, Fast, cached-input, and long-context token rates for the GPT-5.6 family. Fast mode guide Current request parameters, Priority Processing compatibility, billing, ramp-rate behavior, and regional caveats.

Related model routes

GPT-5.2 Review the former flagship's documented variants and a safe migration handoff to the current family. DeepSeek V4 Compare another coding and long-context route while keeping provider and Flowith access boundaries separate. Gemini 3.1 Compare a Google model route when the task depends on a different multimodal or provider boundary. Grok 4 Review another reasoning route when live-answer style analysis is a core workflow requirement.

GPT-5.6 questions, answered

GPT-5.6 is OpenAI's model family launched for general availability on July 9, 2026. The family includes Sol, Terra, and Luna tiers for different capability, cost, and throughput roles.
Sol is the flagship tier, Terra is the balanced lower-cost tier, and Luna is the fastest and most affordable tier. Choose by workload requirements rather than treating the family as one interchangeable model.
OpenAI documents gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna. The gpt-5.6 alias routes to gpt-5.6-sol according to the current model guidance.
For Standard API processing with short context, OpenAI lists Terra at $2 per million input tokens, $0.20 cached input, and $12 output. Luna is $0.20 input, $0.02 cached input, and $1.20 output. Long-context, Batch, Flex, Fast, regional, tool, and cache-write charges can differ, so verify the live pricing table before budgeting.
No subscription price change was announced in the July 30 update. OpenAI says ChatGPT and Codex subscription prices and quota budgets remain unchanged, while Terra and Luna use fewer credits on paid subscriptions. Your actual allowance still depends on the live product, plan, and account terms.
OpenAI renamed Priority Processing to Fast mode on July 30, 2026. For GPT-5.6 Sol, OpenAI says Fast mode can be up to 2.5 times faster than Standard at twice the token price. API requests using either the fast or legacy priority service-tier value remain supported, but regional availability and ramp-rate behavior still need verification.
OpenAI says GPT-5.6 is available across ChatGPT, Codex, and the OpenAI API, with plan-dependent access. Those provider surfaces do not automatically establish availability inside Flowith.
This page does not claim that GPT-5.6 is currently available on Flowith. Check the live Flowith workspace model selector and account entitlements before planning a workflow.
Test a migration on representative tasks first. Preserve the endpoint, tools, output contract, reasoning behavior, latency role, and cost role before adopting new GPT-5.6 features.
Start with Sol when quality is the priority, Terra for a balanced everyday role, and Luna for efficient high-volume work. Validate the choice against your own task-success, latency, and cost requirements.