NewHow we catch a seller listing the wrong model, and refund you if one does.

What is the cheapest GPT API?

On November 20, 2026, the lowest-priced GPT model on MartLLM is GPT-OSS 20B at $0.000018 per million input tokens and $0.000089 per million output tokens, 99.9% below its list price. MartLLM serves 36 GPT models through one OpenAI-compatible API; the table lists the price of each. Prices come from sellers on the MartLLM market and change during the day. MartLLM adds no fee to them.

Prices as of November 20, 2026, 11:00 UTC. This page is rebuilt every hour.

Against other providers: Cheapest GPT API: below OpenAI on 10 of 10 models (November 2026)

ModelInput / 1M tokensOutput / 1M tokensBelow list price
GPT-OSS 20B$0.000018$0.00008999.9%
GPT-OSS 120B$0.000069$0.00029799.9%
GPT OSS Safeguard 20B$0.000075$0.000399.9%
GPT OSS Safeguard 120B$0.0015$0.00699%
GPT-6 Luna$0.00299$0.01597%
GPT 5 Nano$0.00966$0.077380.7%
GPT-5.6 Luna$0.02$0.1290%
GPT-4o Mini$0.0261$0.104582.6%
GPT 5.4 Nano$0.0324$0.202783.8%
GPT-6 Sol$0.0414$0.20797.9%
GPT-5.6 Terra$0.0408$0.244998%
OpenAI: GPT-6.1 Terra$0.0461$0.230795.4%
GPT-6.1 Sol$0.0608$0.304297%
GPT 5 Mini$0.0484$0.386980.7%
GPT-6 Luna Pro$0.0798$0.39920.2%
GPT-5.4 Mini$0.073$0.437990.3%
GPT-5.6 Sol$0.1242$0.620896.9%
GPT-5.6 Luna Pro$0.1397$0.837930.2%
GPT-6.2 Sol$0.1595$0.797492%
GPT-4o$0.2163$0.86591.4%

MartLLM adds no fee to these prices · every model and price →

More questions

Can I compare GPT with lower-priced models?

Yes. The same SDK and key can call DeepSeek, GLM, Qwen and other models, so you can run the same workload on several models and compare cost and output.