GPT API key | OpenAI API | Codex | GPT-6 Astra, GPT-6.1 Sol | 50% off + cache

GPT API key | OpenAI API | Codex | GPT-6 Astra, GPT-6.1 Sol | 50% off + cache

apiToken — API access to original GPT models: requests run on OpenAI´s own servers, with no model substitution. 50% off every model, with cached input billed by the official rules. Works with Codex CLI and Cursor out of the box.

Same request format as the official OpenAI API, same models and IDs at half the price — and cached input is another 10–20x cheaper.
━━━━━━━━━━━━━━

💸 50% OFF ALL MODELS
50% off OpenAI´s official pricing — on every model and every request. Exactly 50%, on everything.

Formula: our price = official price × 0.5
For every $1 of official API usage, you pay $0.50.

Prices per 1M tokens (input / output):
• GPT-6 Astra: official $10 / $50 → ours $5 / $25
• GPT-6.1 Sol: official $2 / $10 → ours $1 / $5
• GPT-6 Sol: official $2 / $10 → ours $1 / $5
• GPT-6 Luna: official $0.10 / $0.50 → ours $0.05 / $0.25
• GPT-5.6 Sol: official $4 / $20 (promo until Nov 21, 2026, then $5 / $30) → ours $2 / $10 (then $2.50 / $15)
• GPT-5.6 Terra: official $2 / $12 → ours $1 / $6
• GPT-5.6 Luna: official $0.20 / $1.20 → ours $0.10 / $0.60
• GPT Image 2 / 2.5, image output: official $30 → ours $15

━━━━━━━━━━━━━━

🧠 CACHE — WHERE THE REAL SAVINGS ARE
In Codex and other agents about 90% of input tokens are context re-read from the cache. A cache read costs 10% of regular input or less, and the 50% discount applies to it too.
Agent-session input as a share of the official price:
• here, 50% + cache: 0.5 × (0.9 × 0.1 + 0.1) ≈ 0.1
• 70% off without cache (all input at full rate): 0.3
Bottom line: 70% off without cache ends up about 3x more expensive than our 50% with cache.
Caching follows OpenAI´s official rules. Cached input costs 10% of input, 5% on GPT-6.1 Sol. E.g. GPT-6.1 Sol: $0.10 → ours $0.05 per 1M.
━━━━━━━━━━━━━━

🔎 ORIGINAL MODELS — AND YOU CAN CHECK IT
Anyone can verify:
• Cache: repeat a long request (1024+ tokens) — the second response shows cached_tokens in usage above zero.
• Tokens: input_tokens in the response matches OpenAI´s official count for the same request — a different model has a different tokenizer.
• Model: the model field in the response is exactly the requested ID, e.g. gpt-6.1-sol.
━━━━━━━━━━━━━━

⚡️ NATIVE CODEX AND CURSOR SUPPORT
OpenAI Responses API and Chat Completions: Codex CLI, Cursor, OpenAI SDK and any OpenAI-compatible app.
Just set the base URL and key — tools and code stay the same. Streaming, up to 872K tokens of context.
━━━━━━━━━━━━━━

🔐 SPENDING CONTROL
• Per-key spending limits (daily and monthly), expiry dates and allowed-model lists
• Request history — see exactly what each request cost
━━━━━━━━━━━━━━

🔒 PRIVACY
We do not store the text of your requests and responses and do not use them to train models. Your history keeps only the model, token counts and cost, which billing needs.
━━━━━━━━━━━━━━

✅ WHY apiToken
• Original OpenAI models, no substitution
• Exactly 50% off every GPT model and the cache
• Native Codex CLI support
• No OpenAI account or foreign card needed
• Instant automatic delivery, 24/7

🚀 HOW TO GET STARTED
1. Choose the amount and pay
2. After payment you are taken to apitoken.sale — sign in or sign up and the balance is credited automatically
3. If the page did not open, paste the order´s unique code in your dashboard → "Balance" or at apitoken.sale/redeem
4. Create an API key and add it, together with the API base URL, to Codex or Cursor

Teams and high volumes: custom terms and invoices/paperwork — message us in the order chat.
━━━━━━━━━━━━━━

IMPORTANT: this is a balance top-up for the apiToken service, NOT a ChatGPT Plus or Pro subscription and not a separate OpenAI account. apiToken is an independent service, not affiliated with OpenAI. Models are accessed with an API key through the apiToken service. The same balance works with Claude and Gemini, also at 50% off.
The 50% discount applies to new models too. The current list and prices are in your apiToken dashboard.