# Latency Measurement on Synthetic sequences performance benchmarking Source: [Latency Measurement on Synthetic sequences performance benchmarking]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:48:29.221Z Evaluation focusing on measuring latency, speedup, and sparsity characteristics derived from synthetic sequence performance benchmarks. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Latency Measurement | | Dataset (including any stated split) | Synthetic sequences performance benchmarking | | Metric | Latency (ms) | | Unit | TIME | | Direction | Lower is better | | Results loaded | 11 of 11 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Latency (ms)](): Lower is better; unit: TIME. - [Speedup](): Higher is better; unit: SCALAR. - [Sparsity](): Higher is better; unit: PERCENT. ## Recorded results | Method | Latency (ms) | Result date | Report paper | | --- | --- | --- | --- | | Topological Attention | 0.03 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Flash Attention | 0.04 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Topological Attention | 0.19 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Topological Attention | 2.78 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Flash Attention | 3.27 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Topological Attention | 3.94 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Topological Attention | 7.88 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Topological Attention | 15.76 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Topological Attention | 31.52 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Flash Attention | 156.25 | Not recorded | 69896a2fc97a0b3e3b5dc60c | | Flash Attention | 627.58 | Not recorded | 69896a2fc97a0b3e3b5dc60c | ### Topological Attention Latency (ms): 0.03. Result ID: 6998e92f5698665ec44d9d65. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"1,024","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Flash Attention Latency (ms): 0.04. Result ID: 6998e92f5698665ec44d9d64. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"1,024","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Topological Attention Latency (ms): 0.19. Result ID: 6998e92f5698665ec44d9d67. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"8,192","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Topological Attention Latency (ms): 2.78. Result ID: 6998e92f5698665ec44d9d69. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"65,536","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Flash Attention Latency (ms): 3.27. Result ID: 6998e92f5698665ec44d9d66. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"8,192","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Topological Attention Latency (ms): 3.94. Result ID: 6998e92f5698665ec44d9d6b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"131,072","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Topological Attention Latency (ms): 7.88. Result ID: 6998e92f5698665ec44d9d6c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"262,144","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Topological Attention Latency (ms): 15.76. Result ID: 6998e92f5698665ec44d9d6d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"524,288","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Topological Attention Latency (ms): 31.52. Result ID: 6998e92f5698665ec44d9d6e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"1,048,576","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Flash Attention Latency (ms): 156.25. Result ID: 6998e92f5698665ec44d9d68. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"65,536","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. ### Flash Attention Latency (ms): 627.58. Result ID: 6998e92f5698665ec44d9d6a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seq Length":"131,072","Hardware":"NVIDIA RTX 4090 Laptop (16GB VRAM)","Batch size":1,"Attention config":"8 heads, 64 dimensions per head","Window size":"64 tokens"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.