DeepInfra model pricing.

Prices checked

DeepInfra sells 198 prices on 126 models, from $0.02 per million tokens, checked .

Every model DeepInfra sells

  1. 1

    Llama 3.2 3B Instruct

    Runs it themselves

    $0.02 in · $0.02 out

    $0.02/M tokens
    View
  2. 2

    Llama 3.1 8B Instruct

    Runs it themselves

    $0.02 in · $0.04 out

    $0.03/M tokens
    View
  3. 3

    Llama 3.1 8B Instruct

    Lower precision (fp8)

    $0.02 in · $0.04 out

    $0.03/M tokens
    View
  4. 4

    Mistral Nemo

    Same as maker

    $0.03 in · $0.03 out

    $0.03/M tokens
    View
  5. 5

    Llama 3.1 8B Instruct

    Runs it themselves

    $0.03 in · $0.05 out

    $0.04/M tokens
    View
  6. 6

    Llama 3 8B Lunaris

    Same as maker

    $0.04 in · $0.05 out

    $0.04/M tokens
    View
  7. 7

    $0.05 in · $0.05 out

    $0.05/M tokens
    View
  8. 8

    Qwen2.5 7B Instruct

    Runs it themselves

    $0.04 in · $0.10 out

    $0.06/M tokens
    View
  9. 9

    gpt-oss-20b

    Costs more

    $0.03 in · $0.14 out

    $0.06/M tokens
    View
  10. 10

    gpt-oss-20b

    Costs more

    $0.03 in · $0.14 out

    $0.06/M tokens
    View
  11. 11

    gpt-oss-20b

    Costs more

    $0.03 in · $0.14 out

    $0.06/M tokens
    View
  12. 12

    Mistral Small 3

    Same as maker

    $0.05 in · $0.08 out

    $0.06/M tokens
    View
  13. 13

    Mistral Small 3

    Same as maker

    $0.05 in · $0.08 out

    $0.06/M tokens
    View
  14. 14

    Gemma 3 12B

    Runs it themselves

    $0.04 in · $0.12 out

    $0.06/M tokens
    View
  15. 15

    Gemma 3 4B

    Same as maker

    $0.05 in · $0.10 out

    $0.06/M tokens
    View
  16. 16

    Gemma 3 4B

    Same as maker

    $0.05 in · $0.10 out

    $0.06/M tokens
    View
  17. 17

    Gemma 3 4B

    Same as maker

    $0.05 in · $0.10 out

    $0.06/M tokens
    View
  18. 18

    NVIDIA Nemotron Nano 9B v2

    Runs it themselves

    $0.04 in · $0.16 out

    $0.07/M tokens
    View
  19. 19

    Phi 4

    Runs it themselves

    $0.06 in · $0.11 out

    $0.07/M tokens
    View
  20. 20

    gpt-oss-120b

    Runs it themselves

    $0.04 in · $0.17 out

    $0.07/M tokens
    View
  21. 21

    gpt-oss-120b

    Runs it themselves

    $0.04 in · $0.17 out

    $0.07/M tokens
    View
  22. 22

    gpt-oss-120b

    Runs it themselves

    $0.04 in · $0.17 out

    $0.07/M tokens
    View
  23. 23

    Gemma 3 12B

    Same as maker

    $0.05 in · $0.15 out

    $0.08/M tokens
    View
  24. 24

    Gemma 3 12B

    Same as maker

    $0.05 in · $0.15 out

    $0.08/M tokens
    View
  25. 25

    Nemotron 3.5 Lightning

    Runs it themselves

    $0.06 in · $0.16 out

    $0.08/M tokens
    View

Not ranked

Below the maker's own price with nothing published to account for it, or from a catalog we could not read reliably. Why

  1. Mistral Nemo

    Not ranked

    $0.02 in · $0.03 out

    $0.02/M tokens
    View
  2. $0.06 in · $0.06 out

    $0.06/M tokens
    View
  3. Qwen3.8 Flash

    Not ranked

    $0.11 in · $0.38 out

    $0.18/M tokens
    View
  4. GLM 5

    Not ranked

    $0.60 in · $2.08 out

    $0.97/M tokens
    View
  5. Qwen3.8 Max

    Not ranked

    $1.65 in · $4.95 out

    $2.48/M tokens
    View

Where to start

Kind

Host

based in US

Free tier

No

every price is paid

Where DeepInfra ranks

What DeepInfra charges per million tokens embedded

Priced on input alone, since an embedding model writes no output, so never held beside a chat model's blended price. Every embedding model

Embedding models, per 1M input tokens

  1. 1

    Bge Base En v1 5

    No maker price

    Beijing Academy of Artificial Intelligence

    0.5¢/M tokens
  2. 2

    BGE M3

    Runs it themselves

    Beijing Academy of Artificial Intelligence

    $0.01/M tokens
  3. 3

    E5 Large v2

    Runs it themselves

    intfloat

    $0.01/M tokens
  4. 4
    $0.01/M tokens
  5. 5
    $0.01/M tokens
  6. 6

    Qwen3 Embedding 0.6B

    Runs it themselves

    Alibaba Cloud

    $0.01/M tokens
  7. 7

    Qwen3 Embedding 8B

    No maker price

    Alibaba Cloud

    $0.01/M tokens
  8. 8

    Qwen3 Embedding 4B

    No maker price

    Alibaba Cloud

    $0.02/M tokens

What DeepInfra charges per image, clip or recording

Priced by the job, so never held beside a price per million tokens. Every image model

Images, per image

  1. 1

    Sdxl Turbo

    No maker price

    Stability AI

    0.02¢per image
  2. 2

    FLUX.1 [schnell]

    No maker price

    Black Forest Labs

    0.05¢per image
  3. 3

    FLUX.1 [dev]

    No maker price

    Black Forest Labs

    0.9¢per image
  4. 4

    FLUX.2 [dev]

    No maker price

    Black Forest Labs

    1¢per image
  5. 5

    FLUX.2 [klein] 4B

    No maker price

    Black Forest Labs

    1.4¢per image
  6. 6

    FLUX.2 [klein] 9B

    No maker price

    Black Forest Labs

    1.5¢per image
  7. 7

    FLUX.2 [pro]

    No maker price

    Black Forest Labs

    1.5¢per image
  8. 8

    Wan 2.6 Image

    No maker price

    Alibaba Cloud

    3¢per image

Video, per 5-second clip

  1. 1

    Pixverse 6

    No maker price

    PixVerse

    $0.23per 5-second clip
  2. 2

    Wan 2.2 T 2v A 14b

    No maker price

    Alibaba Cloud

    $0.38per 5-second clip
  3. 3

    Wan 2.6

    No maker price

    Alibaba Cloud

    $0.50per 5-second clip
  4. 4

    Wan

    No maker price

    Alibaba Cloud

    $0.50per 5-second clip
  5. 5

    Veo 3.1 Fast

    Costs more

    Google

    $0.75per 5-second clip
  6. 6

    Wan v3.0 Video

    No maker price

    Alibaba Cloud

    $1per 5-second clip
  7. 7

    Veo 3.1

    Same as maker

    Google

    $2per 5-second clip

Speech, per 1M characters

  1. 1

    Kokoro 82M

    No maker price

    hexgrad

    $0.99per 1M characters
  2. 2

    Csm 1b

    No maker price

    Sesame

    $7per 1M characters
  3. 3

    Orpheus 3b 0.1 Ft

    No maker price

    Canopy Labs

    $7per 1M characters

Speech, per hour of audio

  1. 1

    Whisper large-v3 turbo

    No maker price

    OpenAI

    1.2¢per hour of audio
  2. 2

    Whisper large-v3

    No maker price

    OpenAI

    2.7¢per hour of audio

Frequently asked

What is DeepInfra?
Serves open-weight models on its own GPUs, often at fp8 or fp4, and publishes the precision per model. It sells 126 of the models we track, and trades as Deep Infra, Inc., based in US.
Is DeepInfra cheaper than buying from the maker?
103 of its 198 prices are below the maker's own price, and 90 of those come with a published reason, so they can lead a ranking. 5 are listed and not ranked, because nothing accounts for the gap.
Does DeepInfra train on what you send it?
DeepInfra does not say whether it trains on what you send it. It does not say how long it keeps requests.
How current are these DeepInfra prices?
They come from its own public catalog, read every 3 days; the reading on this page is from 31 Oct 2026. Prices exclude tax.

Models it sells

Other hosts

Where to go next