Gemini 3.7 Flash vs GPT-5.6 Luna: multimodal or cheapest text

Chatbots, classification, summarization and lightweight agents all make a lot of calls, so price per token and speed matter more than peak intelligence. In that tier, the two models people compare most are Google's Gemini 3.7 Flash and OpenAI's GPT-5.6 Luna.

The short answer: Gemini 3.7 Flash if your pipeline includes audio or video; GPT-5.6 Luna if it's mostly text and you want the lowest bill.

Specs and price

Luna is about a quarter of Gemini's input price and a third of its output price. Luna does charge more once a prompt passes 272K tokens ($0.42 input, $1.90 output per million), which still leaves it well under Gemini. Prices are what a USD account pays today; current figures are on the pricing page.

Where each one is stronger

Gemini 3.7 Flash is built around wide input. Google designed it to take audio and video alongside text and images, so a job like summarizing a recorded meeting or tagging a video library can run on a single model instead of a transcription step plus a text model. Google also tuned it for responsiveness in multi-step agent loops. On FastMetal, image input is verified on the current route and audio input is billed at the same rate as text; try a short video clip before you build on video input.

GPT-5.6 Luna wins on price. For chat, classification and routing, where the model needs to be good enough and the volume is high, that gap adds up fast. It also shares a prompt style with GPT-5.6 Terra and Sol, so prompts you tune on Luna carry over when you step up.

Things to watch

Gemini 3.7 Flash: for text-only bulk work you are paying three to four times Luna's rate for modalities you don't use.

GPT-5.6 Luna: no audio or video input; text and images only. As the light model of its family, it falls behind the larger ones on hard multi-step reasoning. The usual setup is to send only the difficult requests to Terra or Sol.

Both: PDF uploads are switched off on both routes on FastMetal for now. Extract the text, or send pages as images.

Which one to use

• Gemini 3.7 Flash: pipelines with audio or video, multimodal agents, or a stack you want to keep on Google models.
• GPT-5.6 Luna: high-volume text, cost-first chatbots, or a model ladder that stays inside GPT-5.6.

Try both

Same key, same balance; only model changes:

curl https://api.fastmetal.ai/v1/chat/completions \
  -H "Authorization: Bearer $FASTMETAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gemini-3.7-flash", "messages": [{"role": "user", "content": "Hello"}]}'

curl https://api.fastmetal.ai/v1/chat/completions \
  -H "Authorization: Bearer $FASTMETAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-5.6-luna", "messages": [{"role": "user", "content": "Hello"}]}'

Real answers from both models to the same prompts are on the comparison page, and details are on the Gemini 3.7 Flash and GPT-5.6 Luna pages.