Get alerts
All products/LLM APIs/OpenAI API

OpenAI API pricing, price hikes and cheaper alternatives

OpenAI API (openai.com) is a LLM API. Pay-as-you-go per token: each model bills per 1M input and output tokens, with extra rates for cached input, prompts over 272K tokens and some regional processing, plus per-call fees for built-in tool. No free plan.

What OpenAI API costs

  • Top model: $200/mo (GPT-6 Astra at standard short-context rates: 10 x $10 input + 2 x $50 output = $200.00/mo. Flagship proxy: the LMArena rank could not be checked (arena.ai returned 403), so the top OpenAI model may differ.)
  • Cheapest model: $2/mo (gpt-6-luna, the lowest-cost model that is not retiring and supports function calling with a 1,050,000-token context: 10 x $0.10 + 2 x $0.50 = $2.00/mo. Cheaper gpt-5-nano ($1.30/mo) and gpt-4.1-nano ($1.80/mo) are excluded as deprecated, retiring Dec 11, 2026 and Oct 23, 2026.)
  • gpt-oss-120b: not offered (n/a: OpenAI's API pricing lists no per-token gpt-oss-120b, so the point is not offered.)

Plan compared: Pay-as-you-go per token, GPT-6 Astra at standard rates (flagship proxy). Monthly USD cost of one reference workload: 10,000 requests a month, totaling 10M input and 2M output tokens, at standard pay-as-you-go rates with no caching or batch discounts. Top model is the product's highest-ranked model on the LMArena Text leaderboard. Cheapest model is its lowest-cost model with tool calling and a 128K context. gpt-oss-120b compares hosts on the same open model.

What OpenAI API does

Core: Frontier-grade models, OpenAI-compatible endpoint, Hosted agent tools.

Important: Image and PDF input, 1M-token context, Voice and realtime audio, Image and video generation, Prompt caching discounts, Batch and flex pricing, Managed fine-tuning, No-training, zero retention (partly), Compliance and data residency (partly).

Also has: Open-weight models (partly), Embeddings and reranking (partly), Dedicated model endpoints (partly), Flat-rate coding plan (partly).

OpenAI API price hikes

  • Realtime text output rose 50%, to $24 per 1M · May 11, 2026
    GPT-Realtime-2 text output went from $16 to $24 per 1M tokens; audio and image rates held steady and two realtime models were added. This is not a text-model rate, so it does not move the top point.
  • 10% surcharge on GPT-5.4 for in-region processing · Mar 11, 2026
    Keeping data in-region adds a 10% surcharge to GPT-5.4 models; the standard global rates stayed at $2.50 input and $15 output per 1M tokens.
  • Flagship API rates rose 40% with GPT-5.2 · Jan 12, 2026
    Input and output rates rose 40% in the GPT-5.2 listing, which the source says was GPT-5 renamed; the flagship workload went from $32.50 to $45.50 a month. The before rates are inferred from the 40% rise, not stated on the page.

Cheaper alternatives to OpenAI API

At 12M tokens a month on its top model, OpenAI API costs $200/mo. 1 LLM API costs less:

  • Gemini API: $40/mo (Pay-as-you-go Standard paid tier, text only, list prices as of 2026-10-17), covers 85% of OpenAI API; missing or partial: OpenAI-compatible endpoint (partly), Managed fine-tuning, Dedicated model endpoints

OpenAI API alternatives: 3 LLM APIs compared, at every size.

Other kinds of products that cover part of OpenAI API

  • Hugging Face (GPU cloud & model hosting): covers 50% of OpenAI API; missing or partial: Frontier-grade models (partly), Hosted agent tools, Image and PDF input (partly), 1M-token context (partly), Voice and realtime audio (partly)
  • Perplexity (AI assistants): covers 47% of OpenAI API; missing or partial: OpenAI-compatible endpoint (partly), Image and PDF input, 1M-token context (partly), Voice and realtime audio, Image and video generation

Every change we logged

  • Oct 5, 2026 · Price cut: GPT-6.1 Sol cached input was halved; its $2.00 input and $10.00 output rates were unchanged. The model card now shows GPT-6.1 Sol in place of GPT-6 Sol.
  • May 11, 2026 · Price increase: GPT-Realtime-2 text output went from $16 to $24 per 1M tokens; audio and image rates held steady and two realtime models were added. This is not a text-model rate, so it does not move the top point.
  • Apr 27, 2026 · Price cut: GPT-image-2, which replaced GPT-image-1.5, cut image output from $32 to $30 per 1M tokens.
  • Apr 2, 2026 · Price cut: The web search tool price was cut 60%, from $25 to $10 per 1k calls; content tokens from search stay free.
  • Mar 24, 2026 · Price increase: GPT-5 mini was renamed GPT-5.4 mini and its input, cached and output rates all rose; its workload went from $6.50 to $16.50 a month (not the top point). GPT-5.4 nano was added at $0.20 input and $1.25 output.
  • Mar 11, 2026 · Price increase: Keeping data in-region adds a 10% surcharge to GPT-5.4 models; the standard global rates stayed at $2.50 input and $15 output per 1M tokens.
  • Jan 12, 2026 · Price increase: Input and output rates rose 40% in the GPT-5.2 listing, which the source says was GPT-5 renamed; the flagship workload went from $32.50 to $45.50 a month. The before rates are inferred from the 40% rise, not stated on the page.

Sources

Pricing page: openai.com/api/pricing. Prices last checked Oct 16, 2026. Prices are from OpenAI's developer pricing docs and model pages, checked directly; the price-change list comes from PricingSaaS's openai.api log, a secondary summary. LMArena could not be checked (arena.ai returned 403), so the top point is a flagship proxy, and ChatGPT seat plans and Containers billing are left out as not per-token API prices.