AI/ML●●Solid
Doppler.js – WebGPU inference, faster/simpler than transformer.js
Explicit kernel control over TVM-style black boxes, but benchmarks show mixed wins vs Transformers.js.
Big BrainWizardry
clocksmith
303mo ago
Back to browse
WebGPU LLM inference comprehensive benchmark
by yu3zhou4·Apr 6, 2026·2 points·2 comments
AI Analysis
●●SolidBig BrainNiche Gem
Sequential-dispatch methodology corrects 20x overestimation in prior WebGPU benchmarks.
Strengths
- •torch-webgpu backend achieves 11-12% of CUDA performance as out-of-tree PyTorch extension
- •Cross-vendor data spanning NVIDIA, AMD, Apple, Intel is rare and genuinely useful
- •Open source code, benchmarks, and raw data available for verification
Weaknesses
- •11-12% of CUDA performance means not viable for production inference workloads yet
- •Research paper format means less polished DX than dedicated inference tools
Category
Target Audience
ML engineers building browser-based AI, WebGPU developers, performance researchers
Similar To
WebLLM · Transformers.js · MLPerf
Similar Projects
AI/ML●●Solid
AI/ML benchmark for local LLM inference and XGBoost training on GPU/CPU
One-command benchmark suite comparing Ollama and XGBoost performance with a shared Streamlit dashboard.
Solve My ProblemNiche Gem
albedan
2018d ago
AI/ML●●●Banger
LLM Sycophancy Benchmark: Opposite-Narrator Contradictions
Opposite-narrator test catches models agreeing with both sides of same dispute.
Big BrainDark Horse
zone411
302mo ago
AI/ML●●Solid
LLM Debate Benchmark
Side-swapped debate matchups expose model weaknesses standard benchmarks miss.
Big BrainDark Horse
zone411
932mo ago
AI/ML●●Solid
ErrataBench - A Proofreading Benchmark for LLMs
51 models, 1613 runs, $558 spent — finally proofreading benchmarks with real numbers.
Niche GemBig Brain
artursapek
301mo ago
AI/ML●Mid
My "home rig" for iterative attribute-weighted LLM benchmarking
Home rig for attribute-weighted benchmarking lacks the polish of established eval frameworks.
Ship It
yuvalhaim
211mo ago