ModelsPricingDocs

Open models, one meter

Every model on Aperture is open-weight, priced per token, and metered in plain sight. Choosing a model is choosing a price — so the price is on the card, not in a footnote.

live catalog · prices per 1M tokens · licenses shown (they matter)

  • gpt-oss 20B

    OpenAI OSS

    OpenAI's open-weight 20B — quick and capable.

    $0.05/M in$0.20/M out128k ctx
    tools
    Apache-2.0
    Try gpt-oss 20B

  • Gemma 4 31B

    Gemma

    Google's open Gemma 4 instruct, 31B.

    $0.05/M in$0.15/M out256k ctx
    tools
    Gemma
    Try Gemma 4 31B

  • Nemotron 3 Super 120B

    Nemotron

    NVIDIA's large open reasoning model.

    $0.30/M in$0.60/M out256k ctx
    toolsreasoning
    NVIDIA Open Model
    Try Nemotron 3 Super 120B

  • Nemotron 3 Ultra 550B

    Nemotron

    NVIDIA's flagship — 1M context, slower.

    $0.50/M in$1.50/M out1M ctx
    toolsreasoning
    NVIDIA Open Model
    Try Nemotron 3 Ultra 550B

  • North Mini Code

    Cohere

    Cohere's coding-focused open model.

    $0.10/M in$0.30/M out256k ctx
    CC-BY-NC-4.0
    Try North Mini Code

  • Nemotron Nano 9B

    Nemotron

    Small and fast Nemotron Nano.

    $0.02/M in$0.08/M out128k ctx
    NVIDIA Open Model
    Try Nemotron Nano 9B

See the meter run

One capped request against gpt-oss 20B, streamed live, metered at its real per-token prices. No account.

Watch the meter run — no account needed

gpt-oss 20B · capped at 200 tokens · gpt-oss-20b

Start with $0.20 of free credit

Enough for a real conversation on a cheap model. See how billing works.

Get started