The purpose of Instill Credit is easing the adoption of 鈽侊笍 Instill Cloud, minimizing the time required to build and set up a pipeline. After setting up your account, you'll get 10,000 monthly credits for free, which can be spent on the following actions:
- Run your own pipelines, or any public pipeline available on Explore.
- Execute pre-configured AI components without needing to create accounts or API keys on 3rd party services.
We offer different subscription plans for users that need more monthly credits to fulfil their pipeline run and API consumption needs.
#Pipeline Run
Each pipeline run will consume 1 credit per component within the pipeline.
This excludes the trigger and response components which don't consume any credit, as they are the pipeline's input and output interfaces.
#Model Run
The model credit consumption is calculated as follows: active period * credit pricing.
Here, active period refers to the number of seconds the model takes to process a request, while credit pricing varies based on the hardware used by the model.
The table below lists the available hardware options along with their corresponding credit pricing.
Hardware | Credit Price (per second) | Credit Price (per hour) |
---|---|---|
CPU | 1 | 3600 |
NVIDIA Tesla T4 | 2 | 7200 |
NVIDIA L4 | 3 | 10800 |
NVIDIA A100 | 13 | 46800 |
#AI Component
Additionally, Instill AI facilitates out-of-the-box configuration for certain AI components. When such components are created, if the selected task and model are supported, the default configuration will contain the credentials to start using the component.
If you want to override the default configuration and bring your own key, you can update the Connection configuration in your component. Connection fields only take references to secrets in order to minimize the risk of leaking credentials when sharing your pipelines. Therefore, you'll need to create a secret with your API key in your account settings page.
To go back to using the default credentials on your component, you can reference
the ${secrets.INSTILL_SECRET}
in the component configuration. You may use this
reference as well if you want to modify an existing pipeline (e.g. one you
cloned from Explore) and use the default credentials on any supported
component.
The following section details the supported AI Tasks & models and their Instill Credit cost per unit.
#Supported AI Tasks and Models
#LLM
Vendor | Model | Credit Cost per 1,000 Input Tokens | Credit Cost per 1,000 Output Tokens |
---|---|---|---|
OpenAI | OpenAI o1-preview | 150 | 600 |
OpenAI | OpenAI o1-mini | 30 | 120 |
OpenAI | GPT-4o | 50 | 150 |
OpenAI | GPT-4o 2024-08-06 | 25 | 100 |
OpenAI | GPT-4o mini | 1.5 | 6 |
OpenAI | GPT-4 | 300 | 600 |
OpenAI | GPT-4 (32K) | 600 | 1,200 |
OpenAI | GPT-4 (128K) | 100 | 300 |
OpenAI | GPT-4 VISION | 100 | 300 |
OpenAI | GPT-3.5 Turbo | 5 | 15 |
Anthropic | Claude 3.5 Sonnet | 30 | 150 |
Anthropic | Claude 3 Opus | 150 | 750 |
Anthropic | Claude 3 Sonnet | 30 | 150 |
Anthropic | Claude 3 Haiku | 2.5 | 15 |
Cohere | Command-R+ | 30 | 150 |
Cohere | Command-R | 5 | 15 |
Cohere | Command | 10 | 20 |
Cohere | Command-light | 3 | 6 |
Mistral AI | open-mixtral-8x22b | 20 | 60 |
Mistral AI | open-mixtral-8x7b | 7 | 7 |
Mistral AI | open-mistral-7b | 2.5 | 2.5 |
Mistral AI | codestral-latest | 10 | 30 |
Mistral AI | mistral-large-latest | 40 | 120 |
Mistral AI | mistral-small-latest | 10 | 30 |
Fireworks AI | llama-v3p1-405b-instruct | 30 | 30 |
Fireworks AI | llama-v3p1-70b-instruct | 9 | 9 |
Fireworks AI | llama-v3p1-8b-instruct | 2 | 2 |
Fireworks AI | llama-v3-70b-instruct | 9 | 9 |
Fireworks AI | llama-v3-8b-instruct | 2 | 2 |
Fireworks AI | firellava-13b | 2 | 2 |
Fireworks AI | firefunction-v2 | 9 | 9 |
Fireworks AI | deepseek-coder-v2-lite-instruct | 5 | 5 |
Fireworks AI | starcoder-16b | 2 | 2 |
Fireworks AI | starcoder-7b | 2 | 2 |
Fireworks AI | phi-3-vision-128k-instruct | 2 | 2 |
Fireworks AI | qwen2-72b-instruct | 9 | 9 |
Fireworks AI | mythomax-l2-13b | 2 | 2 |
Fireworks AI | yi-large | 30 | 30 |
Groq | llama3-groq-70b-8192-tool-use-preview | 8.9 | 8.9 |
Groq | llama3-groq-8b-8192-tool-use-preview | 1.9 | 1.9 |
Groq | llama3-70b-8192 | 5.9 | 7.9 |
Groq | llama-guard-3-8b | 0 | 0 |
Groq | llama3-8b-8192 | 0.5 | 0.8 |
Groq | mixtral-8x7b-32768 | 2.4 | 2.4 |
Groq | gemma2-9b-it | 2 | 2 |
Groq | gemma-7b-it | 0.7 | 0.7 |
#Text Embeddings
Vendor | Model | Credit Cost per 1,000 Input Tokens |
---|---|---|
OpenAI | Text Embedding 3 Small | 0.2 |
OpenAI | Text Embedding 3 Large | 1.3 |
Cohere | Embed 3 | 1 |
Mistral AI | mistral-embed | 1 |
Fireworks AI | nomic-ai/nomic-embed-text-v1.5 | 0.08 |
Fireworks AI | nomic-ai/nomic-embed-text-v1 | 0.08 |
Fireworks AI | WhereIsAI/UAE-Large-V1 | 0.16 |
Fireworks AI | thenlper/gte-large | 0.16 |
Fireworks AI | thenlper/gte-base | 0.08 |
#Image Generation
Vendor | Model | Image Size | Credit Cost per Image |
---|---|---|---|
OpenAI | DALL路E 3 Standard | 1024x1024 | 400 |
OpenAI | DALL路E 3 Standard | 1024x1792, 1792x1024 | 800 |
OpenAI | DALL路E 3 HD | 1024x1024 | 800 |
OpenAI | DALL路E 3 HD | 1024x1792, 1792x1024 | 1,200 |
Stability AI | SDXL 1.0 | 60 | |
Stability AI | SD 1.6 | 100 |
#Audio Recognition
Vendor | Model | Credit Cost per Second |
---|---|---|
OpenAI | Whisper | 1 |
#Text to Speech
Vendor | Model | Credit Cost per 1M Characters |
---|---|---|
OpenAI | TTS 1 | 150 |
OpenAI | TTS 1 HD | 300 |
#Text Reranking
Vendor | Model | Credit Cost Per 1000 Searches |
---|---|---|
Cohere | Rerank 3 | 20,000 |
#Next Additions
Subscribe to our Newsletter to keep up to date with the latest tasks, models and vendors that are supported by Instill Credit.
We also plan to leverage Instill Credit to ease data storage and model hosting, stay tuned!
#Organization Credit
Team plans also include monthly Instill Credit for organizations, allowing their members to consume the organization's credit instead of their own.
Users can impersonate an organization and consume its Instill Credit by switching namespaces. Any pipeline accessible by the organization (i.e. any public pipeline or any private pipeline owned by the organization) can be triggered using this method.
#How can I get more credit?
The amount of monthly Instill Credit of a user or organization depends on their subscription plan.
Additionally, subscribers can purchase extra credit in case the monthly amount doesn't cover their Instill Credit needs. In contrast to monthly credit, purchased credit won't expire at the end of the billing cycle. For that reason, purchased credit will only be consumed only when the subscription credit has been exhausted.
Our roadmap includes more features to cover more complex Instill Credit use cases, such as Credit Auto-billing, where credit is topped up before it is totally exhausted, keeping production environments safe from any potential downtime.