Best AI Models for Reasoning and Math

Reasoning models trade latency for accuracy on multi-step problems. This ranking weighs benchmark intelligence most heavily, since reasoning quality is what you are paying for.

8 models ranked · updated continuously

How we rank: Ranked by benchmark intelligence on math, logic, and multi-step tasks.

  1. 01

    o5OpenAI

    OpenAI's next reasoning flagship.

    IQ
    1450
    Rating
    4.85 / 5
    Context
    256K
    Price
    $15/M tokens
    License
    Closed
  2. 02

    Gemini 3 Deep ThinkGoogle DeepMind

    Long-horizon reasoning over 2M tokens.

    IQ
    1410
    Rating
    4.70 / 5
    Context
    2M
    Price
    $8/M tokens
    License
    Closed
  3. 03

    DeepSeek R3DeepSeek

    Open-weights reasoning powerhouse.

    IQ
    1395
    Rating
    4.75 / 5
    Context
    128K
    Price
    $0.5/M tokens
    License
    Open Weights
  4. 04

    o5-miniOpenAI

    Affordable o-series reasoner.

    IQ
    1360
    Rating
    4.60 / 5
    Context
    256K
    Price
    $2/M tokens
    License
    Closed
  5. 05

    OpenAI o4OpenAI

    Deep reasoning specialist

    IQ
    1352
    Rating
    4.80 / 5
    Context
    256K
    Price
    $15/M tokens
    License
    proprietary
  6. 06

    QwQ 2Alibaba

    Qwen's reasoning specialist.

    IQ
    1350
    Rating
    4.50 / 5
    Context
    128K
    Price
    $0.4/M tokens
    License
    Open Weights
  7. 07

    DeepSeek R2DeepSeek

    Open frontier reasoning

    IQ
    1342
    Rating
    4.70 / 5
    Context
    256K
    Price
    $0.55/M tokens
    License
    open-weights
  8. 08

    OpenAI o4-miniOpenAI

    Affordable reasoning workhorse

    IQ
    1337
    Rating
    4.60 / 5
    Context
    200K
    Price
    $3/M tokens
    License
    proprietary

More rankings