GPT-6 comes in three tiers. Astra is the frontier model, released on 2026-09-03. Sol and Luna followed on 2026-09-22 as the everyday and the high-volume tiers. From Astra to Luna the list price drops 100x, so choosing the tier moves your bill far more than choosing a provider. This page puts OpenAI list prices next to what APIMODELS charges on each tier, and says plainly where going direct is still the better call.
On APIMODELS, Astra is $2.40 input / $12 output / $0.24 cached input per 1M tokens, 76% below the $10 / $50 list. Sol is $1 / $5 / $0.10, exactly half of its $2 / $10 list. Luna is $0.09 / $0.45 / $0.009, 10% below its $0.10 / $0.50 list. The discount is deliberately not uniform: OpenAI already priced Luna close to the floor, and we would rather state the real per-tier number than advertise one headline figure that only holds at the top.
If you are still on GPT-5.6, the upgrade is also a price cut. OpenAI lists GPT-6 Sol at half of GPT-5.6 Sol ($2 / $10 against $4 / $20), and on APIMODELS GPT-6 Sol at $1 / $5 is cheaper than GPT-5.6 Sol at $1.324 / $6.618. Both keep the same 1,050,000-token context window and 128,000-token output limit.
Long context has its own step, and we pass it through rather than hiding it. When a request carries more than 272,000 input tokens, OpenAI bills the whole request, not just the tokens past the line, at 2x input and 1.5x output. APIMODELS applies the same 2x / 1.5x step on all three tiers.
Reliability, measured rather than claimed: over the last 30 days of external production traffic (internal accounts excluded), Astra served 18,025 calls at a 99.8% success rate, with a median of 16.0 seconds and a p95 of 111.8 seconds. Sol and Luna went live on the platform on 2026-09-23 and have about 150 calls each so far, too few to quote honestly, so we are not quoting them. Failed requests are never billed.
Access is the other half of the comparison. APIMODELS needs no OpenAI account and no organization verification, works from mainland China, bills one USD balance across every model, and takes Stripe, PayPal and Alipay. Both /v1/chat/completions and /v1/responses are supported, so existing OpenAI SDK code only changes its base URL and key.
Price per 1M tokens — OpenAI list vs APIMODELS, verified 2026-09-26
input output cached in vs list
GPT-6 Astra
OpenAI (list) $10.00 $50.00 $1.00
APIMODELS $2.40 $12.00 $0.24 -76%
GPT-6 Sol
OpenAI (list) $2.00 $10.00 $0.20
APIMODELS $1.00 $5.00 $0.10 -50%
GPT-6 Luna
OpenAI (list) $0.10 $0.50 $0.01
APIMODELS $0.09 $0.45 $0.009 -10%
Long context: above 272,000 input tokens the whole request bills at
2x input / 1.5x output, on both sides.
Context window 1,050,000 tokens; max output 128,000 tokens (all tiers).cURL
# Drop-in OpenAI endpoint — only the base URL and key change
curl https://api.apimodels.app/v1/chat/completions \
-H "Authorization: Bearer $APIMODELS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-6-sol",
"messages": [{"role": "user", "content": "Explain prompt caching in two sentences."}]
}'Python
from openai import OpenAI
client = OpenAI(
api_key="YOUR_APIMODELS_KEY",
base_url="https://api.apimodels.app/v1",
)
# gpt-6-astra for the hardest agent work, gpt-6-luna for high-volume jobs
r = client.chat.completions.create(
model="gpt-6-sol",
messages=[{"role": "user", "content": "Summarise this changelog."}],
)
print(r.choices[0].message.content)
print(r.usage) # cached tokens show up here and bill at the cached rateOpenAI list prices per 1M tokens are $10 / $50 for Astra, $2 / $10 for Sol and $0.10 / $0.50 for Luna, with cached input at $1.00, $0.20 and $0.01. On APIMODELS the same tiers are $2.40 / $12, $1 / $5 and $0.09 / $0.45, with cached input at $0.24, $0.10 and $0.009: about 76%, 50% and 10% below list.
Astra for the hardest long-horizon agent and research work. Sol is the everyday default for chat, coding and tool-calling agents, at a fifth of Astra’s list price. Luna is for high-volume, cost-sensitive jobs such as classification, extraction and rewriting, at a twentieth of Sol’s list price. All three sit behind the same OpenAI-compatible endpoint, so switching is a one-string change.
Because OpenAI already prices Luna close to the floor at $0.10 / $0.50 per 1M tokens, so there is far less room underneath it than under Astra’s $10 / $50. If per-token cost is your only criterion for Luna, OpenAI direct is already close; the deciding factors are access and billing, not price.
At the Sol tier, yes. OpenAI lists GPT-6 Sol at $2 / $10, half of GPT-5.6 Sol’s $4 / $20. On APIMODELS, GPT-6 Sol is $1 / $5 against $1.324 / $6.618 for GPT-5.6 Sol. Context window and output limit are the same on both generations.
Above 272,000 input tokens, the whole request is billed at 2x input and 1.5x output. That is OpenAI’s own rule, and APIMODELS applies the same step on all three tiers. Below the line, the normal per-token prices apply.
Sign up at apimodels.app, create an API key, and point the OpenAI SDK at https://api.apimodels.app/v1. No OpenAI account or organization verification is needed, and it works from mainland China. Use the model ids gpt-6-astra, gpt-6-sol or gpt-6-luna; gpt-6 on its own is an alias for gpt-6-sol.
Three honest cases. First, the Batch API: OpenAI discounts asynchronous batch jobs by 50%, which we do not resell, so large offline workloads can be cheaper there. Second, enterprise compliance paperwork and data-residency commitments that only a direct OpenAI or Azure contract provides. Third, priority processing and reserved capacity, which are account-level products rather than per-request options.
