📊 Hurtig beslutningsoversigt
✅ Anbefalet primær
x-ai/grok-4.1-fast
Score 95% • 6.7s median • 0 fejl • OpenRouter
🔄 Anbefalet backup
x-ai/grok-4-fast
Score 95% • 8.2s median • 0 fejl • OpenRouter
🧪 Bedste NanoGPT
nousresearch/hermes-3-llama-3.1-70b
Score 95% • 11.7s median • 0 fejl • NanoGPT
⚡ Hurtigste fallback
openai/gpt-4o-mini
Score 87.5% • 2.2s median • 0 fejl • OpenRouter
🆓 Free-LLM Manager
inclusionai/ling-2.6-1t:free
Score 96.7% • metadata-only collector som default
🚫 Blokeret / Undgå
openai/gpt-5-nano
Score 0.0% • teknisk ok men ingen scoring • undgå
📢 TTS-venlig CoS-opsummering
🚀 Frontier Top-3 — Level 2 resultater
👥 Fuld rolle-matrix
RolleModelScore / StatusMiljø
🆓 Free-LLM Manager

🆓 Free-LLM Manager — inclusionai/ling-2.6-1t:free

Model: inclusionai/ling-2.6-1t:free

Score: 96.7% role-eval

Rate-limits:

20 requests/minut50 requests/dag1000 requests/dag

Default-mode: metadata-only collector

Brug: Kun til metadata-indhentning og rolle-evaluering. Ingen produktionskald uden eksplicit godkendelse.

Evidence: /opt/data/home/hermes-llm-eval/free_llm_manager/results/free_llm_final_report_20260504.txt

📡 OpenRouter Collector / Cron / Skill
💰 Billig-agent rolle-evaluering
🔗 CoS Pipeline — Autonomi & Gates
📜 Autonomi-politik
⚠️ Stop-gates & Risikogates
📋 Alle 20 cheap-agent resultater — synlige fra start
# ▼ Model ▼ Layer ▼ Score ▼ OK kald ▼ Median latens ▼ Fejl ▼ Repair ▼ Status
🎯 Næste handlinger — hvad systemet bør køre
📁 Evidence paths — Summary & rapport artefakter