Model guide · updated 2026

Cheapest AI models (that are still worth using)

Cheap is easy; cheap and capable is the real goal. These are the lowest-cost models by blended price per 1M tokens, filtered to ones that still hold up on the Artificial Analysis intelligence index — so you are not paying less for a model that cannot do the job.

  1. 1

    DeepSeek V4 Flash

    DeepSeekTop pick

    High-volume coding assistants and IDE integrations where LiveCodeBench 91.6% and SWE-bench 79% performance matters at low per-token cost

    $0.18
    Blended / 1M
  2. 2

    GPT OSS 120b

    OpenAI

    Agentic coding and engineering workflows

    $0.26
    Blended / 1M
  3. 3

    Grok-4 Fast Reasoning

    xAI

    Mathematical and STEM problem-solving where benchmark-level analytical depth matters

    $0.28
    Blended / 1M
  4. 4

    Grok-4.1 Fast Reasoning

    xAI

    Production agent pipelines requiring accurate

    $0.28
    Blended / 1M
  5. 5

    Grok-3 Mini

    xAI

    Math, logic, and quantitative reasoning where benchmark accuracy (95.8% AIME) translates to real gains

    $0.35
    Blended / 1M
  6. 6

    GPT OSS 120B Fast

    OpenAI

    Agentic workflows and multi-tool automation requiring reliable function calling and chain-of-thought reasoning

    $0.45
    Blended / 1M
  7. 7

    GPT-5.4 nano

    OpenAI

    Data extraction and document parsing at scale

    $0.46
    Blended / 1M
  8. 8

    MiniMax M2.5

    MiniMax

    High-volume software engineering and code generation across Web

    $0.52
    Blended / 1M

Ranked by blended price (75% input + 25% output) per 1M tokens, ascending, among models scoring 30+ on the Artificial Analysis intelligence index.

Frequently asked questions

It changes as providers cut prices, but open-weight models from DeepSeek, Zhipu (GLM), Alibaba (Qwen) and MiniMax consistently lead on blended price while still scoring well on intelligence. See the ranked list above for the current order.

Not necessarily. The cheapest models are usually smaller or open-weight, so they trade some peak reasoning for huge cost savings. For extraction, classification, chat and most everyday tasks, the cheap tier is more than enough — that is why we filter to models that still clear an intelligence floor.

On just4o.chat every model here is available from one chat, so you can route cheap tasks to a cheap model and hard tasks to a flagship without separate subscriptions or API keys.