GPT-5.6 API — Sol, Terra and Luna Through One Gateway
OpenAI's GPT-5.6 ships as a three-tier family — Sol (flagship), Terra (balanced), Luna (fast) — GA since 2026-07-09 after a limited preview that started June 25. byesu serves all three as gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna through one OpenAI-compatible endpoint, behind one key, without an OpenAI account — pay-as-you-go, no subscription. The same key also covers Claude, Grok, Gemini and more.
What It Is · How Pricing Works
Official OpenAI list pricing per million tokens:
| Tier | Model name on byesu | Positioning | Official list price (in / out) |
|---|---|---|---|
| Sol | gpt-5.6-sol | Flagship — deepest reasoning, ultra mode workhorse | $5 / $30 |
| Terra | gpt-5.6-terra | Balanced — ~97% of Sol's benchmark performance per OpenAI | $2.50 / $15 |
| Luna | gpt-5.6-luna | Fast — high-volume, latency-sensitive workloads | $1 / $6 |
Cache economics (official): cache writes cost 1.25x the input price, cache reads cost 10% of it, and the Batch API bills at half the standard rate. Two features are new to the 5.6 family:
- Ultra mode — the model orchestrates multiple sub-agents in parallel for complex tasks.
- Predictable caching — a prompt prefix is guaranteed to stay cached for at least 30 minutes, with user-defined breakpoints, so cache behavior is something you design instead of guess at.
byesu serves the official models with no quality degradation and bills per token using group multipliers:
| Billing Item | Multiplier |
|---|---|
| Input | 1x |
| Output | 5x |
| Cache hit | 0.1x |
Multipliers are coefficients relative to each model's base price. You pay a fraction of the official list price — the console shows your live billed price next to the official list price, per tier, and it follows automatically whenever multipliers change. Cache hits bill at only 0.1x.
Pick the right group when creating a token
Before calling any GPT-5.6 tier, go to Console → Tokens and create a token in the OpenAI GPT group. Using the wrong group triggers a "no available channel" error. See Choosing a Group for details.
Sol vs Terra vs Luna: which tier?
The tiers share the same API shape and features — the trade is depth vs cost vs latency. Match the workload, not the badge:
| Workload | Pick | Why |
|---|---|---|
| Multi-step agent pipelines, ultra mode orchestration | Sol | Sub-agent coordination benefits most from the deepest tier |
| Hardest reasoning: architecture reviews, research synthesis, tricky migrations | Sol | Output quality dominates cost when a wrong answer is expensive |
| Production coding assistants and day-to-day dev work | Terra | ~97% of Sol's benchmark performance (per OpenAI) at half the list price |
| RAG over a large, stable corpus | Terra + cache breakpoints | Cache reads at 10% of input price make big shared prefixes nearly free to re-send |
| Bulk classification, extraction, summarization | Luna | Per-token cost dominates at volume; quality ceiling rarely binds |
| Latency-sensitive UX: autocomplete, streaming chat UI | Luna | Fastest tier; pair with streaming |
The cache angle is where the tier choice gets interesting. With predictable caching, a RAG system that pins its corpus prefix behind a breakpoint pays the full input rate once per 30-minute window, then 10% of the input price (0.1x on byesu) on every subsequent hit — which can pull Terra's effective per-request cost below Luna's uncached cost for prefix-heavy workloads. Model your prefix-to-suffix ratio before defaulting to the smallest tier.
Start with Terra; escalate specific request types to Sol when evals show the gap matters, and route bulk paths to Luna. Since all three sit behind one byesu key, that routing is a one-string change.
How to Call It
All three tiers use the OpenAI-compatible endpoint — Base URL https://byesu.com/v1, auth header Authorization: Bearer sk-your-token.
curl /v1/chat/completions
curl https://byesu.com/v1/chat/completions \
-H "Authorization: Bearer sk-YOUR_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-5.6-terra",
"messages": [
{"role": "user", "content": "Summarize the trade-offs between REST and gRPC for internal services."}
]
}'Swap "model" to gpt-5.6-sol or gpt-5.6-luna — nothing else changes.
Python (OpenAI SDK)
Existing OpenAI SDK code needs only a base_url and key swap:
from openai import OpenAI
client = OpenAI(
api_key="sk-YOUR_TOKEN",
base_url="https://byesu.com/v1",
)
for tier in ["gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna"]:
resp = client.chat.completions.create(
model=tier,
messages=[
{"role": "user", "content": "Explain what an API gateway does in one sentence."}
],
)
print(tier, "→", resp.choices[0].message.content)An OpenAI-compatible /v1/responses endpoint is also available for clients that require the Responses wire format (e.g. Codex — see Codex CLI Setup).
Key Parameters
| Parameter | Required | Description |
|---|---|---|
model | ✅ | gpt-5.6-sol, gpt-5.6-terra, or gpt-5.6-luna — exactly as shown in the console |
messages | ✅ | Array of conversation messages, each with role and content |
max_tokens | Optional; if omitted, the model decides the reply length | |
temperature | Sampling temperature — lower is more deterministic | |
stream | Set to true for streaming (SSE) responses |
Output tokens bill at 5x, so tier choice and reply length drive most of the spend; cache hits bill at only 0.1x.
Why Use byesu for GPT-5.6
- All three tiers, one endpoint, no OpenAI account —
gpt-5.6-sol,gpt-5.6-terra, andgpt-5.6-lunabehind a singlesk-token. - Official models, no quality degradation — requests reach the real GPT-5.6 family, including ultra mode and predictable caching behavior.
- Transparent usage-based billing — public multipliers (input 1x / output 5x / cache hit 0.1x); the console shows your live billed price next to the official list price for each tier.
- One key, many vendors — the same token also calls Claude, Grok, and Gemini, so cross-vendor evals are a model-string change.
- Pay-as-you-go, no subscription — top up with USDT, Alipay or WeChat Pay and pay only for tokens you use.
Related Links
- Quick Start (3 steps) — create a token · set the address · pick a model
- Choosing a Group — why the OpenAI GPT group matters
- Console Sign-up / Create Token — get an
sk-token and see live per-tier pricing - Full Documentation — 6 languages, covering text / image / video access
FAQ
How do I get GPT-5.6 API access?
Sign up at the byesu console, create a token in the OpenAI GPT group, and call gpt-5.6-sol, gpt-5.6-terra, or gpt-5.6-luna through the OpenAI-compatible /v1/chat/completions endpoint. All three tiers sit behind one key, no OpenAI account required, billed pay-as-you-go with no subscription.
What is the difference between GPT-5.6 Sol, Terra and Luna?
Sol is the flagship (official list price $5 in / $30 out per million tokens), Terra is the balanced tier ($2.50 / $15) that OpenAI says reaches about 97% of Sol's benchmark performance, and Luna is the fast, high-volume tier ($1 / $6). All three went GA on 2026-07-09 and share ultra mode and predictable caching.
How much does the GPT-5.6 API cost?
Official OpenAI list prices per million tokens: Sol $5 input / $30 output, Terra $2.50 / $15, Luna $1 / $6. Cache writes cost 1.25x the input price, cache reads 10% of it, and the Batch API bills at half the standard rate. On byesu, billing is per token with transparent group multipliers (input 1x, output 5x, cache hit 0.1x). You pay a fraction of the official list price — the console shows your live billed price next to the official list price.
Can I use GPT-5.6 without an OpenAI account?
Yes. byesu is an AI API gateway: one sk- token gives you all three GPT-5.6 tiers plus Claude, Grok, Gemini and more through the same OpenAI-compatible endpoint. Top up with USDT, Alipay or WeChat Pay and pay only for the tokens you use.
What are ultra mode and predictable caching in GPT-5.6?
Ultra mode lets the model orchestrate multiple sub-agents in parallel for complex tasks. Predictable caching guarantees that a prompt prefix stays cached for at least 30 minutes and lets you place user-defined cache breakpoints, so cache hit rates become something you can engineer rather than hope for.
What does a "no available channel" error mean when calling GPT-5.6?
It usually means the token was created in the wrong group. In the byesu console, create the token in the OpenAI GPT group, and confirm the model name is spelled exactly gpt-5.6-sol, gpt-5.6-terra, or gpt-5.6-luna as shown in the console model list.