How much does GPT-5.6 cost? Sol, Terra, and Luna pricing (July 2026)

GPT-5.6 costs between $1 and $5 per million input tokens depending on the tier you choose. OpenAI ships GPT-5.6 in three tiers: Sol at $5 input / $30 output per million tokens, Terra at $2.50 / $15, and Luna at $1 / $6. Sol is the flagship reasoning tier, Terra is the balanced production tier priced at roughly half of GPT-5.5, and Luna is the cheap high-volume tier. GPT-5.6 launched in limited preview on June 26, 2026 through the API and Codex, with general availability promised in the coming weeks. For documentation work the tier you need depends less on the sticker price than on the context you feed the model, which is where Falconer helps: it grounds each run in your real codebase so a cheaper tier can do the job.

TLDR

• GPT-5.6 Sol costs $5 input / $30 output per 1M tokens, matching GPT-5.5 on price.
• GPT-5.6 Terra costs $2.50 input / $15 output per 1M tokens, positioned as GPT-5.5-competitive at half the cost.
• GPT-5.6 Luna costs $1 input / $6 output per 1M tokens, the cheapest tier for high-volume work.
• GPT-5.6 launched in limited preview June 26, 2026, accessed through an OpenAI account representative, not a public waitlist.
• For documentation workloads, input price dominates because a from-scratch docs run reads far more tokens than it writes.

GPT-5.6 pricing by tier

| Tier | Input $/1M | Output $/1M | Best for | | --- | --- | --- | --- | | Sol | $5 | $30 | Complex reasoning, long-horizon agentic work, coding | | Terra | $2.50 | $15 | Everyday production traffic | | Luna | $1 | $6 | High-volume, latency-sensitive apps |

Sol is the flagship tier, built for tasks where correctness matters more than cost. Terra is the balanced middle, positioned by OpenAI as competitive with GPT-5.5 at roughly half the price. Luna is the fastest and most affordable tier, aimed at chatbots, classification, and real-time use.

What does GPT-5.6 cost for a real documentation task?

For documentation work the input price matters most, because generating a docs set from a codebase reads far more tokens than it writes. A run that ingests a large repository and emits a compact docs set will be dominated by the input rate, so Luna at $1 per million input is the cheapest way to run high volume, while Sol at $5 buys the strongest reasoning for tangled code. The bigger lever is context: the less the model has to rediscover on each run, the cheaper the tier you can get away with.

So which tier is cheaper for documentation?

At these rates, Sol is the safer bet than a pricier flagship like GPT-5.5 Pro for writing and updating docs. But you can likely run most doc work on Terra, or even Luna, given the right context. Frontier tiers burn tokens and rack up cost precisely when they lack context: they re-read the codebase, chase dead ends, and over-reason to fill the gaps. Feed a model the right context and a cheaper tier clears the bar. This is what Falconer solves. It grounds every run in your actual code and prior docs, so you can drop to a cheaper tier without losing quality.

How does GPT-5.6 pricing compare to GPT-5.5?

GPT-5.5 costs $5 input / $30 output per million tokens on the standard tier, the same as GPT-5.6 Sol. That makes the comparison simple: Sol matches GPT-5.5 on price while raising the reasoning ceiling, Terra undercuts it at $2.50 / $15 for similar everyday performance, and Luna is far cheaper for high-volume work. GPT-5.5 also moves to a long-context schedule of $10 input / $45 output above roughly 270K tokens, so large-context jobs cost more than the headline rate.

How do you access GPT-5.6?

During the preview there is no public waitlist or self-service signup. Access goes through an OpenAI account representative, and the model is available via the API and Codex to vetted partners. General availability is promised in the coming weeks with no firm date, so if you are planning a documentation workflow around GPT-5.6, confirm your access path before committing. If you are running large jobs, batching requests can halve your per-token cost.

Planning a docs pipeline on GPT-5.6? Falconer grounds each run in your codebase and keeps the output in sync as code changes, so you can run docs on a cheaper tier and trust they stay accurate. It also answers questions right where your team already works.

Run GPT-5.6 docs on a cheaper tier with Falconer

The reason docs runs get expensive is missing context, not the model's rate. Falconer fixes that so you can pick a cheaper tier without losing quality:

• Grounded in your codebase. Every run reads your actual code and prior docs, so the model doesn't re-derive context or over-reason to fill gaps.

• Docs that stay current. Falconer pushes updates back to the source as code changes, so your knowledge base doesn't rot between runs.

• One source of truth. Agents and teammates work from the same accurate docs, not scattered across a dozen tools.

New to the idea? Start with Falconer's guide to technical documentation for engineering teams, then explore Falconer.

FAQ

How much does GPT-5.6 cost per million tokens?

Sol costs $5 input / $30 output, Terra costs $2.50 / $15, and Luna costs $1 / $6 per million tokens. The tier you pick sets the price, and input versus output mix drives your real bill. For documentation work, which reads far more than it writes, the input rate matters most, so run the numbers on your actual workload rather than the headline figure.

What is the difference between Sol, Terra, and Luna?

Sol is the flagship reasoning tier for complex, long-horizon, or tangled coding work where correctness beats cost. Terra is the balanced production tier at roughly half of GPT-5.5's cost, tuned for everyday traffic. Luna is the cheapest, fastest tier for high-volume, latency-sensitive work like chat, classification, and real-time apps. The right pick depends on the task, not a single "best" tier. Falconer lets you route each doc to the tier that fits, sending tangled code to Sol and bulk runs to Luna from one place.

Is GPT-5.6 more expensive than GPT-5.5?

No on two of three tiers. GPT-5.5 costs $5 / $30, which matches Sol but is more than Terra ($2.50 / $15) and Luna ($1 / $6). If you are on GPT-5.5 today, moving everyday traffic to Terra roughly halves your cost with comparable performance, and high-volume jobs on Luna cost a fraction of the GPT-5.5 rate.

Is GPT-5.6 available to the public?

As of late June 2026 it was in limited preview through the API and Codex, with access through an OpenAI account representative rather than a public waitlist. General availability was promised in the coming weeks with no firm date, so confirm your access path before building a workflow around it.

Which GPT-5.6 tier is cheapest for generating documentation?

Luna, at $1 input per million tokens. Because a from-scratch docs run is input-heavy, the input rate drives most of the cost. But cheapest per token is not always cheapest per usable doc: if a lower tier produces docs you have to redo, the rework can cost more than paying for Sol once. Giving the model the right context is what lets a cheaper tier clear the bar, which is where Falconer helps.

How do I keep GPT-5.6 docs accurate over time?

Generating docs once is the easy part; keeping them current as code changes is where most setups rot. Falconer regenerates and pushes updates back to the source as your codebase evolves, so your docs stay grounded in real code instead of drifting out of date between runs.