Warble AI Literacy & MLOps Studio
MLOps / LLMOps Lab & Tech-Business Intelligence
“AI Literacy — It's Never Too Late. Build, Retrospect and Momentum.” Master tiny open LLMs and Kaggle SaaS business datasets.
Tiny Models Registry8 Available
SmolLM2-135M-Instruct
135M
On-device micro-quizzes, intent routing, hint generation
VRAM: 450 MBSpeed: 285 tok/sContext: 8k
SmolLM2-360M-Instruct
360M
Best tiny tutor backbone, interactive coding walkthroughs
VRAM: 950 MBSpeed: 192 tok/sContext: 8k
SmolLM2-1.7B-Instruct
1.7B
Strong conceptual explanations, syllabus drafting
VRAM: 3.6 GBSpeed: 94 tok/sContext: 8k
SmolLM3-3B
3B
Tool calling, 128k context multi-document RAG, multilingual reasoning
VRAM: 6.2 GBSpeed: 62 tok/sContext: 128k
TinyLlama-1.1B-Chat-v1.0
1.1B
Easiest 1B model to fine-tune with community LoRA adapters
VRAM: 2.2 GBSpeed: 145 tok/sContext: 2k
Gemma 3 270M IT
270M
Ultra-small instruct model for instant classroom routing
VRAM: 650 MBSpeed: 240 tok/sContext: 4k
Cognita TinyLlama (CS Teaching)
1.1B LoRA
CS pedagogy dialogues, guided Socratic debugging
VRAM: 2.4 GBSpeed: 138 tok/sContext: 2k
Omni-Edu-4B (Pedagogy & Diagnosis)
4B
K-12 & Higher-Ed learning diagnosis, Socratic scaffolding
VRAM: 8.4 GBSpeed: 48 tok/sContext: 16k
Inference Arena: SmolLM2-360M-Instruct
Ready for Low-VRAM Inference
Grounding: FineWeb-Edu + Stack-Edu • License: Apache 2.0
Attach Socratic LoRA Adapter
Forces conversational scaffolding without revealing raw answers first.
Estimated Latency: ~120ms