FREELINE free AI models, checked live
All models

Free on NVIDIA Build

NVIDIA Build gives free API endpoints for some models. It needs an NVIDIA account and phone verification. Free endpoints come with rate limits, and NVIDIA can retire them: one on this list, GLM 5.3, is already set to retire around 29 Oct 2026.

Checked by hand on 24 Oct 2026. NVIDIA lists 45 free endpoints in total. These are the ones worth using for chat and code.

  1. Nemotron 3 Ultra 550B

    NVIDIA

    Agentic reasoning and coding. The biggest free one.

  2. Nemotron 3.5 Super VL 120B

    NVIDIA

    Reasoning and vision, for agentic work and coding.

  3. Nemotron 3.5 Lightning 30B

    NVIDIA

    A fast agentic model.

  4. Step 5 Preview

    StepFun

    About 600B and multimodal, with reasoning.

  5. Muse Glimmer 30B

    Meta

    Multimodal reasoning with tool use.

  6. Gemma 4 31B IT

    Google

    Reasoning for coding and agents.

  7. DiffusionGemma 26B

    Google

    A diffusion text model, so replies come back very fast.

  8. Laguna XS 2.1

    Poolside

    33B, built for agentic coding.

  9. GPT-OSS 20B

    OpenAI

    A small reasoning model.

  10. MiniMax M3.1 Flash

    MiniMax

    Also has a Freeline page with every route on record.

  11. GLM 5.3 Flash

    Z.ai

    Also has a Freeline page with every route on record.

  12. GLM 5.3

    Z.ai

    The full model. Its free endpoint retires around 29 Oct 2026.

  13. DeepSeek V4.1 Flash

    DeepSeek

    The base model. The 1017 build is not listed separately.

NVIDIA Build needs an NVIDIA account with a verified phone number. Free endpoints are rate limited, so a busy day shows up as slow replies rather than a bill.