OpenAI ChatGPT API Pricing Calculator - All Text/Chat Models

Batch Tier (50% Discount)

Process non-urgent workloads within 24 hours at half price. Example: GPT-5.6 Sol drops to $2.00/$10.00 per M tokens (vs $4/$20 standard). Perfect for data processing, content generation, and analysis tasks that aren't time-sensitive.

Flex Tier (Variable Latency)

Lower cost with variable latency on the GPT-6, GPT-5.6, GPT-5.5, and GPT-5.4 families, including the Pro variants. Automatically routes requests during lower-demand periods. Good middle ground between Batch and Standard for non-critical applications.

Standard Tier (Default)

Default tier with balanced cost and performance. Most common choice for production applications. Consistent latency and throughput. All models available. Rate limits rise with your usage tier, which upgrades as total credit purchases grow.

Fast Tier (2x Cost)

Formerly called Priority. Costs about 2x Standard for faster processing. Best for user-facing applications that need quick responses. Available on the GPT-6 and GPT-5.6 families, GPT-5.5, GPT-5.4, GPT-5.4 Mini, and GPT-5.3 Codex (not Nano or the Pro variants).

Ultrafast Tier (6x Cost)

The fastest processing OpenAI offers, at 6x Standard. Only available on GPT-6 Astra ($60/$300 per M tokens) and GPT-6.1 Sol ($12/$60). Worth it only when response speed directly drives revenue or user retention.

Prompt Caching (90-95% Off Reads)

Cache frequently used prompts, system messages, or documents. Cached input is billed at 10% of the standard rate on the GPT-6, GPT-5.6, GPT-5.5, and GPT-5.4 families, and 5% on GPT-6.1 Sol (e.g. GPT-5.6 Sol input $4.00/M → cached $0.40/M). On GPT-5.6 and later, the first write of a prefix costs 1.25x input and the cache lasts at least 30 minutes after its last use. GPT-5.5 typically keeps it around 30 minutes and up to 24 hours. Earlier models keep it in memory for about 5-10 minutes of inactivity (up to an hour), or up to 24 hours on the models that support 24h retention. Works for chatbots with consistent system prompts, RAG applications, and repeated context.

Fine-Tuning (Winding Down)

OpenAI is winding down the fine-tuning platform as of May 2026. It's no longer open to new users. Existing fine-tuning users can still create training jobs on GPT-4.1, GPT-4.1-mini (SFT/DPO) and o4-mini (RFT). Fine-tuned models remain available for inference until their base models are deprecated. The current GPT-5.6, GPT-5.5, and GPT-5.4 families are not available for fine-tuning.

Understanding Reasoning Tokens

Reasoning-enabled models (GPT-6 family, GPT-5.6 family, GPT-5.5, GPT-5.5 Pro, GPT-5.4 Pro, GPT-5.3 Codex) generate hidden 'thinking' tokens that are billed as output but not returned in the response. These tokens occupy context window space and can significantly increase costs. A 500-token visible response may consume 2000+ total tokens. Use the chat-latest alias or GPT-5.4 Mini/Nano when you don't need extended reasoning.

Prepaid Credits System

Buy credits upfront with no monthly fees. Credits expire after 12 months. Set auto-reload thresholds to never run out. Volume discounts available for enterprise. No setup fees or subscription costs.

Usage Tracking & Limits

Real-time usage monitoring in dashboard. Set spending limits and alerts. Failed requests aren't charged. Rate limits increase automatically with your usage tier (Build, Launch, Grow), based on total credit purchases.