llm-semantic-router/Vela-1.0-Omni-Nano
Document complete Nano multilingual evaluation and current ranks
Document source-declared training exposure for benchmark references
Refresh general ranks with current Mini text mode
Clarify complete benchmark ranks and size–quality comparisons
Improve Omni Nano audio embeddings and complete benchmark rankings
Refresh Mini reference scores and complete-panel rankings
Label familiar reference models in complete-benchmark comparisons
Add complete-benchmark rankings and size–quality comparisons
Show only current Nano task-level Pareto frontiers
Add measured original small standard benchmark comparisons
Update Omni size and quality comparisons
Compare Omni scores with the original small model
Reduce Omni Nano size with equivalent embeddings
Improve Omni Nano embeddings with complete benchmark results
Update size-quality comparisons with reproduced F2 results
Display all four Pareto figures in a two-column model-card grid
Improve image classification, clustering and cross-modal retrieval; update scores and figures
Add size-quality Pareto plots and simplify model card
Simplify evaluation comparisons and benchmark tables
Align complete English benchmark comparison
Add complete audio embedding benchmark results
Update the Mini standard benchmark comparison
Add complete MTEB English v2 evaluation
Improve multimodal alignment and publish complete benchmark scores
Enable 512-token text inputs and update verified benchmark scores
Clarify task-specific standard evaluation algorithms
Add standard text, image and audio benchmark results
Add class-balanced F1 scores to the evaluation matrix
Clarify score comparisons
Improve multimodal embeddings
Use concise Vela Omni model names
Release Vela Omni Nano with multimodal embeddings and evaluation scores
initial commit