Sol, Terra or Luna?
The GPT-5.6 tier selection guide
As of 2026-10-07 the split across the three GPT-5.6 tiers is unchanged on OpenAI's side: Terra for daily work, Sol only when the task is actually hard, and Luna as the volume tier with the lowest official list price. QCode does not currently support gpt-6-luna or gpt-5.6-luna (not supported on QCode); please use another GPT model — for volume work, gpt-5.6-terra. OpenAI list prices: Sol $4/$20 (OpenAI states the promotional price runs until 2026-11-30), Terra $2/$12. Note that OpenAI has since released GPT-6 Astra (2026-09-03), GPT-6 Sol and GPT-6 Luna (2026-09-22) and GPT-6.1 Sol (2026-09-29); its models page now features GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna.
Switch models on the same endpoint and measure the cost per token yourself; check /models for callable models.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Luna is now $0.20/$1.20 (was $1/$6) and Terra $2/$12 (was $2.50/$15). Sol stayed at $5/$30 as of 2026-08-02 (the 2026-09-12 archive of the official pricing page shows $4/$20, promotional until 2026-11-30). Codex/ChatGPT Work usage counting for Luna/Terra reflects the lower rates; plan stickers and quotas were not cut.
The three tiers, side by side
Figures from OpenAI's official release and docs, first verified 2026-07-10. The Sol and Terra columns of the price and cached-read rows were re-checked on OpenAI's official model pages as of 2026-10-13: Sol $4 / $20, cached input $0.40 (OpenAI states that the promotional price for Sol runs until 2026-11-30); Terra $2 / $12, cached input $0.20.
| Metric | Sol | Terra | Luna |
|---|---|---|---|
| Positioning | Flagship — for the hardest problems | Balanced everyday tier — about half the price | Fastest and cheapest |
| Terminal-Bench 2.1 | 88.8% | 84.3% | 82.5% |
| Price (input / output, per 1M tokens) | $4 / $20 | $2 / $12 | $0.20 / $1.20 |
| Cached reads (−90%) | $0.40 | $0.20 | $0.02 |
| API context / max output | 1.05M / 128K | 1.05M / 128K | 1.05M / 128K |
| Speed | Up to ~750 tokens/s on Cerebras from July | Standard serving speed | Fastest of the three |
Note: Sol Ultra is Sol's high-compute mode (91.9% on Terminal-Bench 2.1), not a separate price tier.
Pick a tier by scenario
Choose Sol: the hardest engineering and reasoning work
Cross-repo refactors, long agent runs, hard debug. Sol is 88.8% on Terminal-Bench 2.1. OpenAI’s @thsottiaux, 2026-07-29 (fetched 2026-08-30): Sol was using Codex limits faster than expected; more willing to work longer, make additional tool calls, and launch subagents; High on Sol can use more tokens than High on GPT-5.5; limits were reset for all ChatGPT Work / Codex users; typical Sol use around 18% longer. Freeze the task scope before making it the default.
Choose Terra: the everyday default
Above-GPT-5.5 quality at about half of Sol's price (Terminal-Bench 84.3% vs GPT-5.5's 83.4%). For day-to-day feature work, code review and test generation, Terra is the best value default.
Choose Luna: high-volume, low-latency, batch
Completions, commit messages, bulk jobs, light CI. The official API list price, re-checked on that tier's model page on 2026-10-13, is $0.20 input and $1.20 output per 1M tokens — the lowest of the three tiers. On OpenAI's Codex pricing page Luna bills 5 credits per 1M input against 100 for Sol (fetched 2026-08-30, same figures re-checked 2026-10-13). QCode does not currently support gpt-6-luna or gpt-5.6-luna (not supported on QCode); for high-frequency light tasks please use another GPT model — for volume work, gpt-5.6-terra.
Cost examples
A typical agentic coding session (800K input + 60K output tokens): about $2.32 on Terra, about $5.80 on Sol, and about $0.23 on Luna at short-context list rates as of 2026-08-02 (illustrative).
Switch per request — no commitment
The tiers are three model ids: gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna. Switch any time with /model in Codex CLI, or per request in the API. On QCode, as of 2026-10-07, gpt-5.6-sol and gpt-5.6-terra share one key and are billed by token (per-model rates on /models); QCode does not currently support gpt-6-luna or gpt-5.6-luna (not supported on QCode), so please use another GPT model — for volume work, gpt-5.6-terra.
Tier selection FAQ
Not sure which tier? What should the default be?
Make Terra your default: better than GPT-5.5 at half of Sol's price. Escalate to Sol when Terra gets stuck, and keep Terra for high-frequency light tasks — QCode does not currently support gpt-6-luna or gpt-5.6-luna (not supported on QCode). This Terra-first strategy is the cost optimum for most teams.
Do the tiers share the same context window and output limit?
Yes. OpenAI's docs confirm all three share the 1.05M-token API context window, 128K max output, the same February 16, 2026 knowledge cutoff and the same reasoning-effort options (none through max). Clients may hand out a lower session cap than that spec; the widely quoted 400K is third-party arithmetic (a 272K input budget plus 128K output), not a figure in OpenAI's model page.
What is Sol Ultra? Is it billed separately?
Sol Ultra is Sol's high-compute mode (91.9% on Terminal-Bench 2.1), not a separate price tier. It burns more reasoning tokens, so a task costs more at the same rates. When you see 91.9% quoted, note that it is the Ultra-mode score.
How are the tiers billed on QCode?
As of 2026-10-07, gpt-5.6-sol and gpt-5.6-terra share one QCode key and are billed by token; per-model rates are on /models. QCode does not currently support gpt-6-luna or gpt-5.6-luna (not supported on QCode); please use another GPT model, such as gpt-6.1-sol, gpt-6-sol or gpt-5.6-terra. Switching tiers needs no account changes — just use a different model id.
Sol vs Terra vs Luna in detail (checked 2026-10-15)
The one-line difference: which earlier tier each one maps to
Sol, Terra and Luna differ in capability and price, not in context length. As of 2026-10-15, OpenAI's official model pages position them like this: GPT-5.6 Sol is the GPT-5.6 flagship for complex professional work and roughly corresponds to the unsuffixed tier of earlier GPT-5 families; GPT-5.6 Terra balances intelligence and cost and roughly corresponds to the earlier mini tier; GPT-5.6 Luna is built for cost-sensitive, high-volume workloads and roughly corresponds to the earlier nano tier.
The three tiers also expose the same API surface. According to OpenAI's model pages as of 2026-10-15, Sol, Terra and Luna all support Chat Completions, Responses and Batch, none of them supports Realtime, Assistants or fine-tuning, and they share the same Responses API tools, including web_search, file_search, code_interpreter, computer_use and mcp. At the API level, what differs is the rate limit: on the Build tier, Sol and Terra each allow 5,000 requests and 1,000,000 tokens per minute, while Luna allows 5,000 requests and 2,000,000 tokens; on the Grow tier, Luna goes up to 30,000 requests and 180,000,000 tokens per minute.
They also launched together: OpenAI's API changelog released Sol, Terra and Luna in one 2026-07-09 entry, and the gpt-5.6 alias points to gpt-5.6-sol.
- Sol: the flagship, successor to the unsuffixed tier, for the hardest work
- Terra: the balanced tier, successor to mini, a good everyday default
- Luna: the cost-first tier, successor to nano, for frequent light tasks
How much more Sol costs than Terra, converted from the official price table
OpenAI's official API prices per million tokens (fetched 2026-10-15): Sol is $4 input, $0.40 cached input and $20 output, and OpenAI states that Sol's promotional pricing runs until 2026-11-30; Terra is $2 input, $0.20 cached input and $12 output. On both, cache writes are billed at 1.25x the uncached input rate.
A worked example converted from OpenAI's price table: one request with 100K input and 10K output tokens and no caching costs 0.1 × $4 + 0.01 × $20 = $0.60 on Sol and 0.1 × $2 + 0.01 × $12 = $0.32 on Terra, so Terra comes to about 53% of Sol.
Above 272K input tokens the math changes: OpenAI's rule prices the full request at 2x input and 1.5x output, on all three tiers. For long-context jobs, first check whether the input really needs to exceed 272K.
Choosing in Codex: credit usage and OpenAI's current advice
The Codex pricing page (fetched 2026-10-15) meters usage in credits: per million input tokens GPT-5.6 Sol uses 100 credits and Terra 50; per million output tokens Sol uses 500 and Terra 300, and Luna is the lowest of the three. OpenAI also notes that a typical GPT-5.6 Sol task may use 2-15 credits, and that the GPT-5.6 Sol, Terra and Luna rates remain unchanged.
OpenAI's Codex changelog of 2026-07-31 gave a mapping when GPT-5.4 left ChatGPT-signed-in Codex: replace gpt-5.4 with gpt-5.6-terra and gpt-5.4-mini with GPT-5.6 Luna. (As of 2026-10-15, the Codex models page instead recommends gpt-6-sol as the replacement for gpt-5.4.)
As of 2026-10-15 the Codex models page recommends GPT-6.1 Sol for complex coding and agentic workflows and GPT-6 Luna for focused, repeatable tasks, while stating that GPT-5.6 Sol, Terra and Luna remain available during the rollout. In the CLI, switch with /model or start with codex --model gpt-5.6-terra.
Using the tiers on QCode
QCode's docs recommend a one-click setup script for Codex CLI: run curl -fsSL https://qcode.cc/install/codex.sh | bash in a terminal (on Windows, irm https://qcode.cc/install/codex.ps1 | iex). It installs the CLI, writes the ~/.codex config and verifies connectivity. Then set model to gpt-5.6-terra in config.toml as your everyday default and switch to gpt-5.6-sol for hard problems; high-frequency light tasks you would have given to Luna can also start on gpt-5.6-terra. The docs also state that Codex and Claude Code share the same QCode plan quota and use the same API key.
Endpoints, as in QCode's docs: for Codex CLI, add a custom provider in ~/.codex/config.toml with base_url https://api.qcode.cc/openai and wire_api responses; for Chat Completions clients such as the official OpenAI SDK, set base_url to https://api.qcode.cc/openai/v1. GPT models cannot use the Anthropic endpoint.
Pay with Alipay, WeChat Pay, a card (Visa / Mastercard), Apple Pay, Google Pay or crypto (multiple coins and networks); activation is self-service right after top-up. Billed per token; per-model rates are listed on /models. For GPT models, when a single request's input (including cached tokens) exceeds 272K tokens, the whole request is billed at the long-context rate: input and cached-read prices ×2, output ×1.5, matching OpenAI's official rule.
More questions on Sol, Terra and Luna
How does API pricing change when moving from GPT-5.4 to GPT-5.6 Terra?
It goes down. Per OpenAI's official model pages (captured 2026-10-15), per million tokens GPT-5.4 costs $2.50 input, $0.25 cached input and $15 output, while GPT-5.6 Terra costs $2, $0.20 and $12. Codex's 2026-07-31 mapping also replaced gpt-5.4 with gpt-5.6-terra. If you were on GPT-5.4 mini (official price $0.75 input, $4.50 output), Terra costs more per token, so estimate against your own usage before switching. On QCode you can call gpt-5.6-terra directly; billed per token, per-model rates are listed on /models.
Should I pick Sol or Terra in Codex?
Start daily coding on Terra: Codex counts 50 credits per million input tokens on Terra versus 100 on Sol. Switch to Sol for cross-repo refactors, long agent runs and hard debugging, using /model or codex --model in the CLI.
Were the three tiers released at the same time?
Yes. OpenAI's API changelog released GPT-5.6 Sol, Terra and Luna in a single 2026-07-09 entry, with no staggered launch.
How long does Sol's promotional price last?
OpenAI states that GPT-5.6 Sol's promotional pricing ($4 input, $20 output per million tokens) is available until 2026-11-30; after that, check OpenAI's pricing page.
Now that GPT-6 is out, should I still use the GPT-5.6 tiers?
You can keep using them: as of 2026-10-15 all three remain in OpenAI's model catalog and the deprecations page lists no shutdown date for them. For new projects, evaluate the GPT-6 series directly; OpenAI says GPT-6 Sol and GPT-6 Luna come at lower token prices than their GPT-5.6 predecessors.
How do I pay and get started with GPT-5.6 on QCode?
Sign up for QCode and pay on the top-up page with Alipay, WeChat Pay, a card (Visa / Mastercard), Apple Pay, Google Pay or crypto (multiple coins and networks). After topping up, buy a plan on qcode.cc/subscribe (this step is paid from your balance), then copy your API key, which starts with cr_, from the dashboard into Codex CLI or your SDK. Everything is self-service and takes a few minutes; billed per token, per-model rates are listed on /models.
GPT-5.6 Sol and Terra are callable on QCode
gpt-5.6-sol / terra share one key with the Claude lineup (QCode does not currently support Luna). Plans from $9/mo; the key goes live the moment payment clears.