⚖️ Free2AITools Nexus Index
The Comprehensive Impact Index for Open-Source AI
The S&P 500 of Open-Source AI
🎯 Our Mission
FNI (Free2AITools Nexus Index) is not just another leaderboard. It's an objective, transparent, and verifiable comprehensive evaluation index. We believe: Credibility = Explainability. Only when we can clearly tell users "why it's ranked #1" can we be considered authoritative.
📐 The Formula
S - Semantic (35%)
Query-time relevance
- • Vector similarity matching
- • AI-powered relevance reranking
- • Context-aware scoring
A - Authority (25%)
Ecosystem gravity
- • Knowledge mesh centrality
- • Cross-entity citations
- • Source credibility
P - Popularity (15%)
Community recognition
- • Downloads & likes
- • GitHub stars
- • Log-scaled metrics
R - Recency (15%)
Freshness & vitality
- • Exponential time decay
- • Type-specific half-lives
- • Staleness detection
Q - Quality (10%)
Completeness & usability
- • README depth
- • Metadata richness
- • Runtime compatibility
🏛️ Three Pillars of Fairness
📜 Forensic Data Traceability
All data has a complete audit trail (source_trail). When you question rankings, we can show raw data snapshots, collection timestamps, and content hashes. Data is traceable, conclusions are verifiable.
⚖️ Radical Neutrality
The FNI scoring system (L5 Analyst) and commercial recommendation system (L6 Merchant) are physically separated. We explicitly prohibit "pay-to-rank." Commercial recommendations are based on technical matching, not bidding.
🔮 Holistic Perspective
We don't just look at benchmarks. A model with high benchmark scores but terrible documentation will be appropriately penalized in FNI. We focus on real-world usability, not just ideal laboratory metrics.
🛡️ Anti-Manipulation Mechanisms
We detect anomalous behavior through multi-dimensional cross-validation:
- ❌ Anomalous Growth Detection: Alerts triggered when 7-day growth exceeds 10x the average
- ❌ Ratio Anomaly Detection: Download/like ratios outside reasonable ranges
- ❌ Content Match Detection: High popularity but no substantial content
Anomalous models are flagged for manual review and their scores are appropriately adjusted.
📋 Version History
V1.0 (2025) — Freshness-Novelty Index
Original four-factor formula (P.V.C.U): Popularity, Velocity, Context, Uniqueness. Single-source (HuggingFace), no semantic component.
V2.0 (2026) — Free2AITools Nexus Index Current
Five-factor S.A.P.R.Q formula. Multi-source aggregation across 7+ platforms. 464K+ entities. Agent structured tags for machine-readable model selection.
🤖 Agent Structured Tags
Beyond the five FNI factors, V2.0 provides structured metadata for AI agent model selection:
ollama_compatible Has GGUF quantization — runnable via ollama run
can_run_local Locally runnable: ≤13B parameters + GGUF available
license_type Classified: permissive, copyleft, non-commercial, or unknown
hosted_on Hosting providers: Replicate, Together, HF Inference (daily updated)
These tags power the select_model API — agents can filter by hardware constraints, license, and deployment target.
🔓 Open Source Commitment
FNI's calculation logic is completely open source:
// github.com/mosesy5688-cell/ai-nexus — full source available Anyone can audit our algorithm and propose improvements. Transparency is the foundation of trust.
Public Trust Is Our Currency
Explainability Is Our Moat
Fair · Transparent · Explainable
Explore Models →