Skip to content
HALOWERK

Denkwerk

throughput estimate

Roofline-style estimate with all assumptions in the result: each decode step reads the active weights once (parameters × bits per weight / 8) plus the KV cache of every sequence (same KV formula as vram-estimate), so t_memory = bytes / (bandwidth × memory_efficiency); it needs 2 × active parameters × batch FLOP, so t_compute = FLOP / (dense FP16 TFLOPS × compute_efficiency); the slower one bounds the step. Prefill tokens/s = TFLOPS × efficiency / (2 × active parameters), time to first token.

POST https://denk.halowerk.com/v1/throughput-estimate

Inputs

Fields of the published input schema
FieldTypeRequiredDescription
hf_modelstringNoHugging Face model id; config.json and the safetensors parameter count are read from the Hub.
params_bnumberNoTotal parameters in billions (overrides the Hub value).
n_layersintegerNoNumber of transformer layers.
n_headsintegerNoAttention (query) heads.
n_kv_headsintegerNoKey/value heads (GQA); defaults to n_heads.
head_dimintegerNoDimension per head; defaults to hidden_size / n_heads.
hidden_sizeintegerNoModel width.
vocab_sizeintegerNoVocabulary size (for the compute buffer).
active_params_bnumberNoActive parameters per token for mixture-of-experts models.
quantstringNo
bits_per_weightnumberNo
gpustringNo
memory_bandwidth_gbsnumberNoCustom GPU memory bandwidth in GB/s.
fp16_tflopsnumberNoCustom dense FP16 tensor TFLOPS.
fp16_accumulatebooleanNoGeForce cards: use the FP16-accumulate tensor rate instead of FP32 accumulate.
n_gpusintegerNo
context_tokensintegerNo
batchintegerNo
prompt_tokensintegerNo
output_tokensintegerNo
kv_cache_typestringNo
memory_efficiencynumberNo
compute_efficiencynumberNo
Show full input schema
{
  "type": "object",
  "additionalProperties": false,
  "properties": {
    "hf_model": {
      "type": "string",
      "pattern": "^[A-Za-z0-9][A-Za-z0-9._-]{0,95}/[A-Za-z0-9][A-Za-z0-9._-]{0,95}$",
      "description": "Hugging Face model id; config.json and the safetensors parameter count are read from the Hub."
    },
    "params_b": {
      "type": "number",
      "exclusiveMinimum": 0,
      "maximum": 5000,
      "description": "Total parameters in billions (overrides the Hub value)."
    },
    "n_layers": {
      "type": "integer",
      "minimum": 1,
      "maximum": 1000,
      "description": "Number of transformer layers."
    },
    "n_heads": {
      "type": "integer",
      "minimum": 1,
      "maximum": 1024,
      "description": "Attention (query) heads."
    },
    "n_kv_heads": {
      "type": "integer",
      "minimum": 1,
      "maximum": 1024,
      "description": "Key/value heads (GQA); defaults to n_heads."
    },
    "head_dim": {
      "type": "integer",
      "minimum": 1,
      "maximum": 4096,
      "description": "Dimension per head; defaults to hidden_size / n_heads."
    },
    "hidden_size": {
      "type": "integer",
      "minimum": 1,
      "maximum": 131072,
      "description": "Model width."
    },
    "vocab_size": {
      "type": "integer",
      "minimum": 1,
      "maximum": 2000000,
      "description": "Vocabulary size (for the compute buffer)."
    },
    "active_params_b": {
      "type": "number",
      "exclusiveMinimum": 0,
      "maximum": 5000,
      "description": "Active parameters per token for mixture-of-experts models."
    },
    "quant": {
      "type": "string",
      "enum": [
        "F32",
        "F16",
        "BF16",
        "Q8_0",
        "Q6_K",
        "Q5_K_M",
        "Q5_K_S",
        "Q5_1",
        "Q5_0",
        "Q4_K_M",
        "Q4_K_S",
        "Q4_1",
        "Q4_0",
        "Q3_K_L",
        "Q3_K_M",
        "Q3_K_S",
        "Q2_K",
        "Q2_K_S",
        "IQ4_NL",
        "IQ4_XS",
        "IQ3_M",
        "IQ3_S",
        "IQ3_XS",
        "IQ3_XXS",
        "IQ2_M",
        "IQ2_S",
        "IQ2_XS",
        "IQ2_XXS",
        "IQ1_M",
        "IQ1_S",
        "FP32",
        "FP16",
        "FP8",
        "INT8",
        "INT4"
      ],
      "default": "F16"
    },
    "bits_per_weight": {
      "type": "number",
      "minimum": 1,
      "maximum": 32
    },
    "gpu": {
      "type": "string",
      "enum": [
        "h100-sxm",
        "h100-nvl",
        "h200-sxm",
        "h200-nvl",
        "a100-80gb-sxm",
        "a100-80gb-pcie",
        "a100-40gb-sxm",
        "a100-40gb-pcie",
        "l40s",
        "l40",
        "l4",
        "a10",
        "t4",
        "rtx-a6000",
        "rtx-4090",
        "rtx-3090",
        "apple-m2-ultra",
        "apple-m4-max",
        "apple-m4-pro"
      ]
    },
    "memory_bandwidth_gbs": {
      "type": "number",
      "exclusiveMinimum": 0,
      "maximum": 100000,
      "description": "Custom GPU memory bandwidth in GB/s."
    },
    "fp16_tflops": {
      "type": "number",
      "exclusiveMinimum": 0,
      "maximum": 100000,
      "description": "Custom dense FP16 tensor TFLOPS."
    },
    "fp16_accumulate": {
      "type": "boolean",
      "default": false,
      "description": "GeForce cards: use the FP16-accumulate tensor rate instead of FP32 accumulate."
    },
    "n_gpus": {
      "type": "integer",
      "minimum": 1,
      "maximum": 64,
      "default": 1
    },
    "context_tokens": {
      "type": "integer",
      "minimum": 1,
      "maximum": 4194304,
      "default": 4096
    },
    "batch": {
      "type": "integer",
      "minimum": 1,
      "maximum": 4096,
      "default": 1
    },
    "prompt_tokens": {
      "type": "integer",
      "minimum": 0,
      "maximum": 4194304,
      "default": 1000
    },
    "output_tokens": {
      "type": "integer",
      "minimum": 0,
      "maximum": 1000000,
      "default": 256
    },
    "kv_cache_type": {
      "type": "string",
      "enum": [
        "f32",
        "f16",
        "bf16",
        "fp8",
        "q8_0",
        "q5_1",
        "q5_0",
        "q4_1",
        "q4_0",
        "iq4_nl"
      ],
      "default": "f16"
    },
    "memory_efficiency": {
      "type": "number",
      "exclusiveMinimum": 0,
      "maximum": 1,
      "default": 0.7
    },
    "compute_efficiency": {
      "type": "number",
      "exclusiveMinimum": 0,
      "maximum": 1,
      "default": 0.5
    }
  }
}

Open original schema

Published input example

{
  "params_b": 8.03,
  "n_layers": 32,
  "n_kv_heads": 8,
  "head_dim": 128,
  "quant": "Q4_K_M",
  "gpu": "rtx-4090"
}

Result & documentation

Machine-readable documentation of this hub

Source of this page: the published HALOWERK catalog. Updated: 2026-11-17 UTC.

More endpoints in this hub

All endpoints in this hub →