Written by Bitslix
Thoughts on agentic coding, open source, and building software that lasts.
Language
Getting Started with BLXBench: Benchmark AI Models in 5 Minutes
Learn how to install, configure, and run your first benchmark with BLXBench - the interactive AI model benchmarking platform from Bitslix.
Understanding BLXBench: How We Benchmark AI Models Fairly and Transparently
Deep dive into BLXBench's benchmark methodology, test design, and scoring system for reliable AI model evaluation.
Compositional Meta-Learning for Mitigating Task Heterogeneity in Physics-Informed Neural Networks
A curated insight from https://rss.arxiv.org/rss/cs.AI
BLXBench 1.0.0 released – What’s changed since 0.7
BLXBench reaches General Availability with version 1.0.0. This release consolidates the work since the 0.7 line, adds first‑class LM Studio and Roblox OpenGameEval support, introduces fair per‑model quotas, and tightens the evidence chain with suite labels and manifest seals.
Compositional Meta-Learning for Mitigating Task Heterogeneity in Physics-Informed Neural Networks
arXiv:2604.26999v1 Announce Type: new Abstract: Physics-informed neural networks (PINNs) approximate solutions of partial differential equations (PDEs) by embedding physical laws into the loss fun...
BLXBench 0.8.5: Cleaner HTML Output for More Reliable AI Model Benchmarks
The latest BLXBench release fixes HTML artifact issues in visual previews, making AI model benchmarking workflows smoother and more reliable.
BLXBench TUI and Shell UX Improvements in v0.6.7 and v0.6.8
Recent BLXBench updates improve terminal workflows with clearer TUI feedback, better shell interaction, and more reliable reporting.
BLXBench is live — a community leaderboard for AI models
BLXBench combines reproducible local benchmark runs with a public leaderboard for scores, latency, cost, and task domains.
BLXRouter is live — a catalogue for free AI models
We’re launching BLXRouter: a public index of zero-cost API models, open models you can self-host, and providers with free tiers — updated regularly, no sign-up required.