Tokenomics

Benchmarking suite for OpenAI-compatible inference servers. Measures throughput, latency, and steady-state performance.

Install

pip install tokenomics

From source

git clone https://github.com/tugot17/tokenomics.git
cd tokenomics
uv venv --python 3.12 --seed && source .venv/bin/activate
uv pip install -e .

Completion Benchmark

Sends chat completion requests to any OpenAI-compatible server and records per-request and system-wide metrics. Requests are non-streaming by default (max throughput); --stream adds TTFT and per-token metrics.

Every run sweeps concurrency and writes one result JSON per sweep point. Two execution modes, mutually exclusive:

• --max-concurrency 1,2,4,… — sustained: holds concurrency constant (realistic production numbers).
• --batch-sizes 1,2,4,… — burst: fires each batch at once (peak throughput).

tokenomics completion --model your-model \
  --scenario "D(1024,256)" \
  --max-concurrency 1,2,4,8,16,32,64,128 \
  --results-dir results/

Building requests

Three ways to produce request content, all swept over concurrency the same way:

Mode	Enable with	Text	Use for
Synthetic	--scenario	random dataset snippets padded to a token budget	throughput curves at controlled input/output lengths
Dataset replay	--replay-dataset	each dataset row, verbatim	real examples; comparing datasets
Images (VL)	--num-images	fixed filler (or --scenario) + synthetic images	vision inference speed

Synthetic (--scenario)

A scenario sets input/output token counts; the prompt is built by concatenating random dataset snippets (with replacement) until it reaches the input budget.

Pattern	Example	Description
D(in,out)	D(100,50)	Fixed token counts
N(mu,sigma)/(mu,sigma)	N(100,50)/(50,0)	Normal distribution
U(min,max)/(min,max)	U(50,150)/(20,80)	Uniform distribution

The snippet source defaults to a bundled AIME dataset; override with --dataset-config (see Dataset config).

Dataset replay (--replay-dataset)

Sends each dataset row verbatim as one request and walks the dataset in order, instead of synthesizing prompts. The prompt set is pinned (same rows, same order across every concurrency level and run), so results line up example by example.

tokenomics completion --model your-model \
  --replay-dataset --dataset-config examples/dataset_configs/humaneval.json \
  --max-concurrency 1,2,4,8,16,32 --max-tokens 1024 \
  --results-dir results/humaneval/

• Sustained mode only; --scenario is ignored.
• --num-prompts N caps the walk to the first N rows (default: whole dataset).
• List prompt columns (MT-Bench prompt, Arena-Hard turns) are reduced to the first turn — MT-Bench runs single-turn.

Bundled configs under examples/dataset_configs/:

Config	Dataset	Domain
gsm8k.json	openai/gsm8k main (test)	grade-school math
math500.json	HuggingFaceH4/MATH-500 (test)	competition math (MATH)
aime25.json	math-ai/aime25 (test)	competition math (AIME 2025)
mbpp.json	google-research-datasets/mbpp sanitized (test)	code generation
humaneval.json	openai/openai_humaneval (test)	code completion
lcb.json	livecodebench/code_generation_lite release_v1 (test)	code generation (LiveCodeBench)
mtbench.json	HuggingFaceH4/mt_bench_prompts (train)	multi-turn chat (first turn)
alpaca.json	tatsu-lab/alpaca_eval (eval)	instruction following (AlpacaEval)
arena_hard.json	lmarena-ai/arena-hard-auto-v0.1 (train)	hard instruction following (Arena-Hard)

Vision replay: a config with an image_column (in addition to prompt_column) replays real image+text examples — each row's image(s) are attached to the request. This is the realistic VL workload (real prompts + real images, generating to natural EOS), e.g. for speculative-decoding evaluation. Bundled vision configs:

Config	Dataset	Domain
chartqa.json	lmms-lab/ChartQA (test)	chart question answering
vqa_rad.json	flaviagiammarino/vqa-rad (test)	radiology visual QA
docvqa.json	lmms-lab/DocVQA (validation)	document-image QA

tokenomics completion --model your-vl-model \
  --replay-dataset --dataset-config examples/dataset_configs/chartqa.json \
  --max-concurrency 1,2,4,8,16 --results-dir results/chartqa/

Images (--num-images)

Attach images to any run — sent as OpenAI content parts (image_url base64 data: URIs, accepted by SGLang and vLLM) — turning it into a VL benchmark. Metrics and plotting are unchanged.

tokenomics completion --model your-vl-model \
  --num-images 5 --image-size 512x512 \
  --max-concurrency 1,2,4,8,16 \
  --results-dir results/vl_512x5/

Image runs default to a short workload (the images dominate, the text is padding): a fixed filler of --input-tokens (default 32) and --max-tokens 32 output.

• --image-size is N (square) or WxH (e.g. 1024x768, lowercase x); --input-tokens 0 = images only.
• Synthetic images are random-noise PNGs, seeded per request → unique (defeats the server's prefix/multimodal caches) yet reproducible. Noise is nearly incompressible (~MBs at 1024×1024), so keep size/count modest or the payload dominates.
• --input-tokens and --scenario are mutually exclusive; pass --scenario to put images on dataset-driven text instead.
• Sweep size/count/length by looping the command (one --results-dir each) and overlaying with plot-completion.

Dataset config

A JSON file with a source and (usually) a prompt_column. File paths are resolved relative to the config file.

{ "source": { "type": "huggingface", "path": "openai/gsm8k",
              "huggingface_kwargs": { "name": "main", "split": "test" } },
  "prompt_column": "question" }

source.type is huggingface, file (.txt/.csv/.json), or aime (bundled shortcut). Add an image_column (alongside prompt_column) to replay a vision dataset — its images (embedded PIL, {bytes/path}, or file paths) are encoded and attached per request. See examples/dataset_configs/ for more.

Output length & reproducibility

--ignore-eos makes every request generate exactly --max-tokens (EOS ignored), fixing output length so throughput isn't skewed by content-dependent token counts. Add it when comparing harnesses, servers, or configs; omit it for realistic, content-driven lengths. Supported by SGLang and vLLM — a no-op on servers that ignore the field.

--max-tokens defaults to 4096 (32 for image runs) and --temperature to 0.7. For fully reproducible runs, use --temperature 0 with --ignore-eos.

Key Options

Flag	Description
--model	Model name (required)
--scenario	Traffic pattern (required unless --replay-dataset or --num-images)
--api-base	Server URL (default: http://localhost:8000/v1)
--max-concurrency	Sustained mode sweep points
--batch-sizes	Burst mode sweep points
--num-prompts	Prompts per sweep point in sustained mode
--num-runs	Runs per sweep point (default: 3)
--max-tokens	Max output tokens (default: 4096; 32 for image runs)
--temperature	Sampling temperature (default: 0.7)
--ignore-eos	Generate exactly --max-tokens, ignoring EOS (SGLang/vLLM) — fixes output length for clean comparisons
--stream	Enable SSE streaming for TTFT/per-token metrics
-n	Completions per request (default: 1)
--dataset-config	Path to dataset config (default: bundled AIME)
--replay-dataset	Send each dataset row verbatim (sustained only; ignores --scenario)
--num-images	Attach N synthetic random-noise images per request (0 = text-only)
--image-size	Synthetic image size: N or WxH (default: 512)
--input-tokens	Filler-text length for image runs (default: 32; 0 = images only)
--results-dir	Output directory (one JSON per sweep value)
--lora-strategy	LoRA distribution: single, uniform, zipf, mixed, all-unique
--lora-names	Comma-separated LoRA adapter names

Metrics

Per-request:

• TTFT — time to first token (streaming only)
• Decode throughput — output tokens/s per request (streaming only)
• TPOT — time per output token (streaming only)
• Per-request latency — end-to-end time per request

System-wide:

• End-to-end output throughput — total_output_tokens / wall_time
• Steady-state output throughput — median tok/s across time buckets where the batch is >= 80% full (streaming only)

Plotting

# Compare multiple benchmarks (overlays each results dir as its own line)
tokenomics plot-completion output.png results_dir1/ results_dir2/

Non-streaming (default) produces a 2-panel plot:

Top	Output throughput
Bottom	Per-request latency

Streaming (--stream) produces a 6-panel dashboard:

	Left	Right
Row 1	TTFT (with TTFO overlaid when it diverges, i.e. reasoning models)	Decode throughput per request
Row 2	End-to-end output throughput	Latency breakdown (prefill vs decode)
Row 3	Steady-state output throughput	Time-series token buckets

Embedding Benchmark

Tests concurrent embedding throughput.

tokenomics embedding \
  --model Qwen/Qwen3-Embedding-4B \
  --sequence_lengths "200" \
  --batch_sizes "1,8,16,32,64,128,256,512" \
  --num_runs 3 \
  --results-dir embedding_results/

tokenomics plot-embedding embedding_results/ embedding_plot.png