NAME
    gcloud container ai profiles list - list compatible accelerator profiles

SYNOPSIS
    gcloud container ai profiles list [--format=FORMAT] [--model=MODEL]
        [--model-server=MODEL_SERVER]
        [--model-server-version=MODEL_SERVER_VERSION]
        [--pricing-model=PRICING_MODEL] [--serving-stack=SERVING_STACK]
        [--serving-stack-version=SERVING_STACK_VERSION]
        [--target-cost-per-million-input-tokens=TARGET_COST_PER_MILLION_INPUT_TOKENS]
        [--target-cost-per-million-output-tokens=TARGET_COST_PER_MILLION_OUTPUT_TOKENS]
        [--target-input-length=TARGET_INPUT_LENGTH]
        [--target-itl-milliseconds=TARGET_ITL_MILLISECONDS]
        [--target-ntpot-milliseconds=TARGET_NTPOT_MILLISECONDS]
        [--target-output-length=TARGET_OUTPUT_LENGTH]
        [--target-ttft-milliseconds=TARGET_TTFT_MILLISECONDS]
        [--use-case=USE_CASE] [--filter=EXPRESSION] [--limit=LIMIT]
        [--page-size=PAGE_SIZE] [--sort-by=[FIELD,...]] [--uri]
        [GCLOUD_WIDE_FLAG ...]

DESCRIPTION
    This command lists all supported accelerators with their performance
    details. By default, the supported accelerators are displayed in a table
    format with select information for each accelerator. To see all details,
    use --format=yaml or --format=csvprofile.

    To get supported model, model servers, and model server versions, run
    gcloud container ai profiles models list, gcloud container ai profiles
    model-servers list, and gcloud container ai profiles model-server-versions
    list.

FLAGS
     --format=FORMAT
        The output format. Default is profile, which displays the profile
        information in a table format, including cost conversions. csvprofile
        displays the profile information in a CSV format.Options include
        csvprofile, profile, and yaml.

     --model=MODEL
        The model.

     --model-server=MODEL_SERVER
        The model server.

     --model-server-version=MODEL_SERVER_VERSION
        The model server version. If omitted, the most recently benchmarked
        version is used. Other options include the model server version of a
        profile, or 'all' to return all versions.

     --pricing-model=PRICING_MODEL
        The pricing model to use to calculate token cost. Currently, this
        supports on-demand, spot, 3-years-cud, 1-year-cud

     --serving-stack=SERVING_STACK
        The serving stack to filter profiles by. If not provided, profiles for
        all serving stacks that support the given model and model server will
        be returned.

     --serving-stack-version=SERVING_STACK_VERSION
        The serving stack version. If omitted, the most recently benchmarked
        version is used. Other options include the serving stack version of a
        profile, or 'all' to return all versions.

     --target-cost-per-million-input-tokens=TARGET_COST_PER_MILLION_INPUT_TOKENS
        The target cost per million input tokens to filter profiles by, unit is
        1 USD up to 5 decimal places.

     --target-cost-per-million-output-tokens=TARGET_COST_PER_MILLION_OUTPUT_TOKENS
        The target cost per million output tokens to filter profiles by, unit
        is 1 USD up to 5 decimal places.

     --target-input-length=TARGET_INPUT_LENGTH
        If specified, results will only show profiles that have an input length
        within 20% of the specified one. Only works alongside output length.

     --target-itl-milliseconds=TARGET_ITL_MILLISECONDS
        If specified, results will only show profiles with instance types that
        can meet the latency target and will show their throughput performances
        at the target inter-token latency (ITL).

     --target-ntpot-milliseconds=TARGET_NTPOT_MILLISECONDS
        The target normalized time per output token (NTPOT) in milliseconds.
        NTPOT is measured as the request_latency / output_tokens. If this field
        is set, the command will only return accelerators that can meet the
        target ntpot milliseconds and display their throughput performance at
        the target latency. Otherwise, the command will return all accelerators
        and display their highest throughput performance.

     --target-output-length=TARGET_OUTPUT_LENGTH
        If specified, results will only show profiles that have an output
        length within 20% of the specified one. Only works alongside input
        length.

     --target-ttft-milliseconds=TARGET_TTFT_MILLISECONDS
        The target time to first token (TTFT) in milliseconds. TTFT is measured
        as the request_latency / output_tokens. If this field is set, the
        command will only return profiles that can meet the target ttft
        milliseconds and display their throughput performance at the target
        latency. Otherwise, the command will return all profiles and display
        their highest throughput performance.

     --use-case=USE_CASE
        If specified, results will only show profiles that match the provided
        use case. Options are: Advanced Customer Support, Code Completion, Text
        Summarization, Chatbot (ShareGPT), Text Generation, Deep Research

LIST COMMAND FLAGS
     --filter=EXPRESSION
        Apply a Boolean filter EXPRESSION to each resource item to be listed.
        If the expression evaluates True, then that item is listed. For more
        details and examples of filter expressions, run $ gcloud topic filters.
        This flag interacts with other flags that are applied in this order:
        --flatten, --sort-by, --filter, --limit.

     --limit=LIMIT
        Maximum number of resources to list. The default is unlimited. This
        flag interacts with other flags that are applied in this order:
        --flatten, --sort-by, --filter, --limit.

     --page-size=PAGE_SIZE
        Some services group resource list output into pages. This flag
        specifies the maximum number of resources per page. The default is
        determined by the service if it supports paging, otherwise it is
        unlimited (no paging). Paging may be applied before or after --filter
        and --limit depending on the service.

     --sort-by=[FIELD,...]
        Comma-separated list of resource field key names to sort by. The
        default order is ascending. Prefix a field with ``~'' for descending
        order on that field. This flag interacts with other flags that are
        applied in this order: --flatten, --sort-by, --filter, --limit.

     --uri
        Print a list of resource URIs instead of the default output, and change
        the command output to a list of URIs. If this flag is used with
        --format, the formatting is applied on this URI list. To display URIs
        alongside other keys instead, use the uri() transform.

GCLOUD WIDE FLAGS
    These flags are available to all commands: --access-token-file, --account,
    --billing-project, --configuration, --flags-file, --flatten, --format,
    --help, --impersonate-service-account, --log-http, --project, --quiet,
    --trace-token, --user-output-enabled, --verbosity.

    Run $ gcloud help for details.
