> For the complete documentation index, see [llms.txt](https://docs.algenta.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.algenta.ai/benchmarks.md). # Benchmarks - [Benchmarks](https://docs.algenta.ai/benchmarks/overview.md): How Algenta measures and publishes performance — what each benchmark claims, the baseline it is measured against, and the rules every published number follows. - [Compute kernels](https://docs.algenta.ai/benchmarks/compute-kernels.md): Quantized matrix-vector results — the fused NF4 path measured against expand-then-multiply, with correctness asserted against the reference kernel. - [NVIDIA kernel benchmarks](https://docs.algenta.ai/benchmarks/nvidia-kernel-benchmarks.md): One portable FP32 kernel, unmodified, measured on NVIDIA T4, L4 and A100 — GEMM and saxpy with bit-exact validation; percentages are against spec-sheet peaks. - [NVIDIA LLM throughput](https://docs.algenta.ai/benchmarks/nvidia-llm-throughput.md): Measured LLM inference throughput of the CUDA-native session on NVIDIA A100, H100, H200 and L4 — Qwen3-0.6B, Qwen2.5-7B and Gemma-3-27B prefill and decode against llama.cpp, vLLM, SGLang and TensorRT- - [NVIDIA LLM serving](https://docs.algenta.ai/benchmarks/nvidia-llm-serving.md): Measured serving characteristics of the CUDA-native engine on NVIDIA A100, L4 and H100 — concurrent serving, single-stream decode, batched decode, prefill, cold start, VRAM, determinism; Gemma 3 27B T - [Attention](https://docs.algenta.ai/benchmarks/attention.md): Attention kernel results — tiled attention memory against sequence length, rotary embedding cost, and the correctness checks that ran alongside every timing. - [Normalization](https://docs.algenta.ai/benchmarks/normalization.md): Normalization kernel results — RMSNorm and LayerNorm forward and backward cost, the price of the numerically safe variance, and gradient checks against finite differences. - [Feed-forward](https://docs.algenta.ai/benchmarks/feed-forward.md): Feed-forward and embedding kernel results — gated activation cost across three variants, the two GELU conventions and why they are not interchangeable, and embedding scatter cost. - [Code retrieval quality](https://docs.algenta.ai/benchmarks/code-retrieval.md): Code retrieval quality on CoIR, measured with nDCG@10 against BM25 baselines evaluated on the same harness. - [Dataframe group-by](https://docs.algenta.ai/benchmarks/dataframe-groupby.md): Group-by aggregation on the H2O db-benchmark, 0.5 GB tier, through Algenta's composed route — six of six questions ahead of both DuckDB 1.5.5 and Polars 1.43.1 on the minimum. - [Sort](https://docs.algenta.ai/benchmarks/sort.md): LSD radix sort over Float64 keys against numpy.sort's introsort — 12 of 14 gate-clean cells ahead across three input orders, from 100,000 to 102.4 million elements. - [Top-k selection](https://docs.algenta.ai/benchmarks/top-k.md): Bounded top-k selection over a flat Float64 column, against numpy.argpartition — a real loss at the smallest size tested, and a 5x-163x win at every size above it. - [Graph betweenness centrality](https://docs.algenta.ai/benchmarks/graph-betweenness.md): Brandes betweenness centrality over a CSR adjacency, against networkx — ahead in every one of 6 cells, from 3.0x at 200 nodes to 4.3x at 2,000. - [Survival analysis](https://docs.algenta.ai/benchmarks/survival-analysis.md): Kaplan-Meier and Cox proportional hazards, against lifelines 0.30.0, on two current real-world datasets plus synthetic scaling — wins from 3.0x to 142.9x, one loss disclosed. - [Least-squares accuracy](https://docs.algenta.ai/benchmarks/least-squares-accuracy.md): Least-squares accuracy on the NIST StRD linear-regression suite, measured in correct digits against the certified values, against fourteen reference arms on the same data. - [Forecasting on the joint axis](https://docs.algenta.ai/benchmarks/forecasting-joint-axis.md): Forecast accuracy and latency measured together on all 100,000 M4 series against statsforecast 2.1.1 — equal or better accuracy at 2.42x-3.99x the speed. - [FinQA exact execution](https://docs.algenta.ai/benchmarks/finqa-exact-execution.md): Execution accuracy on all 8,281 gold programs in the real FinQA dataset — Algenta's exact-arithmetic driver against three naive-reimplementation baselines, on the dataset's own exact-match rule. - [LLM inference overview](https://docs.algenta.ai/benchmarks/llm-inference-overview.md): LLM inference on Apple silicon — prefill, single-stream decode, and batched decode for Qwen3-0.6B-Base, measured against llama.cpp and mlx-lm. - [Prefill (token input) throughput](https://docs.algenta.ai/benchmarks/llm-prefill-throughput.md): Prompt prefill (token-input) throughput for Qwen3-0.6B-Base on Apple silicon — a 256-token prompt at ~10,700 tok/s, ahead of llama.cpp's ~10,199, on any prompt length. - [Decode throughput](https://docs.algenta.ai/benchmarks/decode-throughput.md): Decode tokens per second for Qwen3-0.6B-Base on Apple silicon — 324.8 tok/s median vs mlx-lm 281.6 and llama.cpp 254.0, interleaved order-rotated rounds on the same machine. - [Batched decode throughput](https://docs.algenta.ai/benchmarks/batched-decode-throughput.md): Batched (continuous-batching) decode for Qwen3-0.6B-Base on Apple silicon — many sequences per weight sweep. 2.1-2.6x llama.cpp concurrent throughput at B=8-32, token-verified. - [LLM tool-calling uplift](https://docs.algenta.ai/benchmarks/llm-tool-calling-uplift.md): gpt-5-mini answering 16 real questions against a 33,125-row CSV, with and without Algenta's tool-calling access to its own compiled engine — 16 of 16 correct with tools, 1 of 16 without. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.algenta.ai/benchmarks.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.