OpenAI API Pricing (Oct 2026): Every GPT Model per 1M Tokens

1. Official OpenAI API Pricing Table (Verified 2026-10-27)

All OpenAI API token rates below are quoted in USD per 1,000,000 tokens (1M tokens) and verified against OpenAI's official API pricing documentation on 2026-10-27 (asynchronous Batch API jobs are billed under a separate 50%-off schedule). In the official Standard tier, prompts up to 272,000 input tokens use the short-context rates below; once input exceeds 272,000 tokens, the entire request (or session on gpt-5.5) is billed at the > 272K long-context tier (2× input, 2× cached input, and 1.5× output).

Flagship & Latest Models (GPT-6.1, GPT-6, GPT-5.6, GPT-5.5 & GPT-5.4 — Standard Short Context ≤ 272K)

Other GPT, Codex & o-Series Reasoning Models (Standard Tier per 1M Tokens)

---

2. UniRoute vs. Official OpenAI Price Comparison

USD per 1M tokens; official checked: 2026-10-27. UniRoute retail: live gateway snapshot. Saving = 1 − retail / official, rounded to a whole percent; + means more expensive, 0% means equal at displayed precision.Standard and Discounted describe price bands, not upstream providers. Where routes share a band, one complete lowest default public rate set is shown; your key may use another route or account rate. A snapshot is not a frozen billing contract. Comparison covers ordinary standard context, excluding long-context surcharges and priority processing.

Model	Official input	Official output	UniRoute input	UniRoute output	You save (input / output)
gpt-5.5 · Discounted price band · Standard (≤ 272,000 input tokens)	$5.00	$30.00	$1.40	$8.40	72% / 72%
gpt-5.5 · Standard price band · Standard (≤ 272,000 input tokens)	$5.00	$30.00	$5.00	$30.00	0% / 0%
gpt-5.6-terra · Discounted price band · Standard (≤ 272,000 input tokens)	$2.00	$12.00	$0.56	$3.36	72% / 72%
gpt-5.6-terra · Standard price band · Standard (≤ 272,000 input tokens)	$2.00	$12.00	$2.00	$12.00	0% / 0%
gpt-6-astra · Discounted price band · Standard (≤ 272,000 input tokens)	$10.00	$50.00	$2.80	$14.00	72% / 72%
gpt-6-astra · Standard price band · Standard (≤ 272,000 input tokens)	$10.00	$50.00	$10.00	$50.00	0% / 0%
gpt-6-sol · Discounted price band · Standard (≤ 272,000 input tokens)	$2.00	$10.00	$0.56	$2.80	72% / 72%
gpt-6-sol · Standard price band · Standard (≤ 272,000 input tokens)	$2.00	$10.00	$2.00	$10.00	0% / 0%

When you route OpenAI requests through UniRoute (https://api.uniroute.app/v1), you can select between two public channel tiers per API key (see Channel Management for routing and automatic failover configuration):

• Discounted Tier (Discounted): Multi-provider aggregated channel where OpenAI models are available as low as 28% of the official list price (as of 2026-10-27; always refer to the live comparison table above for current real-time rates). Ideal for daily coding agents, CI/CD testing, and cost-sensitive workloads.
• Standard Tier (Standard): Priced at the official baseline rate; choose it if you would rather not be exposed to discounted-channel fluctuations (see Channel Management).
• Per-key switching & automatic failover: You can bind an API key to a specific tier or allow automatic failover so pre-stream errors transparently retry across healthy upstream channels.

---

3. Interactive OpenAI API Cost Calculator

Estimate your daily or monthly spend across official OpenAI rates and UniRoute channels by selecting a model and entering input tokens per request, output tokens per request, and requests per day:

---

4. Detailed Public Route Snapshot (OpenAIRuntimePrices)

The detailed snapshot below displays UniRoute's complete lowest-price rate vector (input, output, cache read, and cache write from the same route) alongside OpenAI's official Standard short-context reference rates:

Default public-route rates · USD per 1,000,000 tokens. “Standard” and “Discounted” describe price bands, not the upstream provider. Where several routes share a band, this table shows one complete lowest-price rate set; your key may use another route or account rate. Rates cover ordinary processing in the standard context tier, excluding long-context surcharges and priority processing.

Current default retail rates from the UniRoute gateway (USD per 1,000,000 tokens)

Model ID	Price band	Input	Output	Cache read	Cache write	Published route max context
gpt-5.5	Discounted	$1.4	$8.4	$0.14	Not published	272,000 tokens
gpt-5.5	Standard	$5	$30	$0.5	Not published	272,000 tokens
gpt-5.6-terra	Discounted	$0.56	$3.36	$0.056	$0.7	272,000 tokens
gpt-5.6-terra	Standard	$2	$12	$0.2	$2.5	272,000 tokens
gpt-6-astra	Discounted	$2.8	$14	$0.28	$3.5	922,000 tokens
gpt-6-astra	Standard	$10	$50	$1	$12.5	922,000 tokens
gpt-6-sol	Discounted	$0.56	$2.8	$0.056	$0.7	922,000 tokens
gpt-6-sol	Standard	$2	$10	$0.2	$2.5	922,000 tokens

Retrieved at 2026-11-08T14:03:39.070341618Z. Rate snapshot: sha256:df6d1d950e90466610e8049d6cd5947d2cc7f57fbfc52beee1b177b6b180a12d. A snapshot identifies the displayed rate set; it is not a frozen billing contract for a future request. Missing (“Not published”) cache rates mean the public feed omits that retail rate—never treat a missing rate as $0 or uncharged. Published route max context (such as 272,000 tokens on current gpt-5.5 and gpt-5.6-terra routes, or 922,000 tokens on gpt-6-sol and gpt-6-astra routes) reflects the current UniRoute public route limit—not OpenAI’s overall model context window (1,050,000 tokens across all four models) and not OpenAI’s 272,000-input-token short-context pricing threshold. Do not extrapolate default standard rates to long-context requests.

OpenAI official reference rates

Official first-party OpenAI Standard short-context (≤ 272,000 input tokens) API reference rates were checked on 2026-10-27 against OpenAI’s API pricing documentation, model catalog, and Prompt Caching guide. These reference rates exclude priority, flex, batch, long-context (> 272,000 input tokens) surcharges, regional data residency, and built-in tool fees. They are separate from the UniRoute retail rates above; a configured channel rate may differ.

OpenAI official reference · USD per 1,000,000 tokens (Standard short context ≤ 272K input tokens)

Model ID	Tier & context scope	Input	Output	Cache read	Cache write
gpt-5.5	Standard (≤ 272,000 input tokens)	$5	$30	$0.5	No extra write fee (billed as ordinary input)
gpt-5.6-terra	Standard (≤ 272,000 input tokens)	$2	$12	$0.2	$2.5
gpt-6-sol	Standard (≤ 272,000 input tokens)	$2	$10	$0.2	$2.5
gpt-6-astra	Standard (≤ 272,000 input tokens)	$10	$50	$1	$12.5

How to Read the Three Pricing Layers

1. OpenAI Official Reference Rates: The first-party Standard API rates published in OpenAI's API pricing documentation. These numbers provide an upstream baseline, not a promise of what UniRoute or a specific gateway route will charge.
2. UniRoute Default Public Retail Rates (Standard and Discounted): Each row in the live UniRoute table above is a complete lowest-price rate vector from a single public route in that price band at the retrieval timestamp—input, output, cache read, and cache write belong to the same route rather than mixing columns across different channels.
3. Your Key's Applicable Route and Account Rate: The public table does not reflect personal account discounts or key-specific routing rules. When you send an authenticated request, billing follows the channel and rate active for your key at request time.

---

5. Choose the Exact GPT Model ID and Check the 272K Context Threshold

API billing always attaches to the exact model string sent in your request body, not a general model family name:

• gpt-5.5 (OpenAI model reference): General-purpose reasoning and multimodal baseline with a 1,050,000 manufacturer context window and 128,000 max output tokens (272,000 max_context_tokens on current UniRoute public routes).
• gpt-5.6-terra (OpenAI model reference): Lower-input-rate reasoning model with a 1,050,000 context window (922,000 max input / 128,000 max output tokens) and explicit cache-write pricing (272,000 max_context_tokens on current UniRoute public routes).
• gpt-6-sol (OpenAI model reference): High-throughput GPT-6 model for daily coding and agent workflows (1,050,000 context window, 922,000 max input, 128,000 max output tokens; 922,000 max_context_tokens on current UniRoute public routes).
• gpt-6-astra (OpenAI model reference): Flagship high-capability GPT-6 model for deep reasoning and complex tasks (1,050,000 context window, 922,000 max input, 128,000 max output tokens; 922,000 max_context_tokens on current UniRoute public routes).

You can inspect model entries in the UniRoute model catalog before sending traffic. Keep three different token numbers separate when estimating request costs:

• OpenAI's manufacturer model window: 1,050,000 total context tokens and 128,000 max output tokens across the flagship models above.
• UniRoute's published route max context (max_context_tokens): Reports the current UniRoute public route limit—not the boundary where short-context pricing ends.
• OpenAI's ≤ 272,000 short-context pricing threshold: In OpenAI's official Standard schedule, base rates apply only when input is 272,000 tokens or fewer. Once input exceeds 272,000 tokens, OpenAI charges the entire request or session at 2× input (and cache rates) and 1.5× output.

---

6. How Prompt Caching and Reasoning Tokens Affect Request Cost

Prompt Caching Reads and Cache Writes

According to OpenAI's Prompt Caching guide and API pricing documentation, prompt caching behavior differs between GPT-5.6 and later (gpt-5.6-terra, gpt-6-sol, gpt-6-astra) and GPT-5.5 (gpt-5.5):

• Official rule for gpt-5.6-terra, gpt-6-sol, and gpt-6-astra (GPT-5.6 and later):
  • Supports both implicit caching and explicit cache control (prompt_cache_options.mode). The minimum cacheable prefix is 1,024 visible input tokens (excluding hidden system tokens).
  • When a response reports total input_tokens, cached hits (cached_tokens), and newly written cache tokens (cache_write_tokens), ordinary uncached input is: ordinary_input_tokens = input_tokens - cached_tokens - cache_write_tokens
  • Multiply ordinary_input_tokens by the ordinary input rate, cached_tokens by the cache-read rate, and cache_write_tokens by the applicable cache-write rate.
• Official rule for gpt-5.5 and gpt-5.4:
  • Supports implicit caching only. On gpt-5.5, implicit breakpoints are spaced at regular 2,048-token intervals, and reported cached_tokens rounds down to a multiple of 128.
  • In OpenAI's official reference schedule, gpt-5.5 and gpt-5.4 do not charge a separate cache-write fee. Ordinary uncached input is input_tokens - cached_tokens.
• How to read missing cache rates in UniRoute's retail table:
  • If a cache rate in the UniRoute table is shown as Not published, treat that retail rate as unknown or not published—never treat a missing retail rate as $0.

Reasoning Tokens Are Already Inside Output Tokens

As documented in OpenAI's Reasoning guide, internal reasoning tokens generated before the visible answer are already included in output_tokens (and reported under reasoning_tokens for visibility) and billed at the output token rate:

• Do not add reasoning_tokens to output_tokens, or you will double-count reasoning output.
• Reasoning tokens still consume output capacity and incur output charges even if a tight max_output_tokens cap truncates the visible response.

Synthetic Cost Example (gpt-6-sol Short Context)

For a hypothetical gpt-6-sol request (≤ 272,000 input tokens) using OpenAI's verified Standard reference rates ($2 input / $0.20 cache read / $2.50 30m cache write / $10 output per 1M tokens):

• Ordinary uncached input: 1,000 tokens × $2 / 1,000,000 = $0.0020
• Cache read: 2,000 tokens × $0.20 / 1,000,000 = $0.0004
• Cache write: 1,000 tokens × $2.50 / 1,000,000 = $0.0025
• Output (already including any reasoning tokens): 500 tokens × $10 / 1,000,000 = $0.0050
• Synthetic official reference token total: $0.0020 + $0.0004 + $0.0025 + $0.0050 = $0.0099 USD

---

7. Frequently Asked Questions (FAQ)

Does an OpenAI API key cost money, and is there a free tier?

Creating an API key on OpenAI Platform or on UniRoute is free, and there is no recurring monthly fee just to hold an API key. API usage is billed strictly pay-as-you-go per 1M tokens consumed. Whether a newly registered first-party OpenAI account receives promotional trial credits depends on OpenAI's current account verification policy, whereas regular production API traffic requires pre-funding an organization balance (or topping up your UniRoute balance).

Are ChatGPT Plus / Pro ($20 / $200/mo) and the OpenAI API billed together?

No. A ChatGPT subscription (such as Plus or Pro under ChatGPT plan rules) and an OpenAI Platform API balance (or a UniRoute account balance) are completely separate products with separate billing:

How much do the Batch API and Prompt Caching save?

• Batch API (50% savings): On OpenAI's official platform, asynchronous Batch API jobs completed within 24 hours receive a 50% discount on both input and output tokens compared to Standard rates (for example, gpt-6-astra drops from $10.00 / $50.00 to $5.00 / $25.00 per 1M tokens, and gpt-6.1-sol drops from $2.00 / $10.00 to $1.00 / $5.00 per 1M tokens).
• Prompt Caching (50% to 95% savings on cached input reads): Reusing identical prompt prefixes reduces input token read costs by 95% on gpt-6.1-sol ($2.00 → $0.10) and gpt-6.1-terra ($1.00 → $0.05), 90% on gpt-6-astra ($10.00 → $1.00), gpt-6-sol ($2.00 → $0.20), gpt-5.5 ($5.00 → $0.50), and gpt-5.4 ($2.50 → $0.25), 75% on gpt-4.1 ($2.00 → $0.50), and 50% on gpt-4o ($2.50 → $1.25).

---

8. Create a Key and Verify Your Own Usage in the UniRoute Console

1. Create and scope a UniRoute API key: Follow the API key creation guide or open API keys in the UniRoute console. Name the key for your client or environment, review its spend limit, channel preference, and allowed models, and save the secret privately when shown.
2. Send one small, nonsensitive test call: Point your OpenAI-compatible client to https://api.uniroute.app/v1 with your exact model ID (gpt-5.5, gpt-5.6-terra, gpt-6-sol, or gpt-6-astra). If you want to connect the official Codex CLI, follow Connect Codex to UniRoute: API Key and Setup Tutorial.
3. Verify the request in your usage list: Open your UniRoute usage records and click the request to open its detail drawer. Check the timestamp, exact model ID, routed channel, outcome, token breakdown (input, output, and cache usage), and recorded actual cost.
4. Related pricing & API guides: See OpenAI Responses API vs. Chat Completions: Migration Guide, CLI vs. API Pricing, Claude Code API Pricing (2026), and DeepSeek API Pricing (2026).