dan-omni-3b-mobile
Ultra-compact variant of dan-omni-3b optimized for strict mobile memory budgets. Smaller context window (2K) for minimal RAM footprint.
Model Description
| Property | Value |
|---|---|
| Base Model | Qwen2.5-3B (further quantized) |
| Fine-tuning | LoRA on mobile-optimized instruction data |
| Context Length | 2048 tokens |
| Parameters | 3B (compressed) |
| File Size | 1.2 GB |
| RAM Usage | ~1.5 GB at inference |
System Prompt
You are dan, a helpful AI assistant for mobile devices. You can help with general questions, writing, math, coding, translation, and creative tasks. Keep answers concise and natural. Don't explain your architecture unless asked.
Benchmarks
Tested on Intel i9-9880H @ 2.30GHz, 16GB RAM, Ollama runtime.
| Category | Avg tok/s | Prompt tok/s | Tokens | Time |
|---|---|---|---|---|
| Reasoning | 9.4 | 37.6 | 170 | 18.1s |
| Coding | 10.2 | 37.6 | 163 | 16.0s |
| Creative Writing | 9.5 | 37.6 | 37 | 3.9s |
| Instruction Following | 9.9 | 37.6 | 52 | 5.3s |
| Math | 10.0 | 37.6 | 127 | 12.7s |
| General Knowledge | 10.1 | 37.6 | 62 | 6.1s |
| Average | 9.8 | 37.6 | 102 | 10.4s |
Comparison vs Competitors
| Model | Size | Speed | Context | RAM |
|---|---|---|---|---|
| dan-omni-3b-mobile | 1.2 GB | 9.8 tok/s | 2K | ~1.5 GB |
| dan-omni-3b | 2.0 GB | 11.3 tok/s | 4K | ~2.5 GB |
| Gemma 4 E2B | ~1.4 GB | ~35 tok/s | 128K | ~2 GB |
| Qwen2.5-1.5B | ~0.95 GB | ~28 tok/s | 32K | ~1.5 GB |
| Command-R7B | ~4.0 GB | ~8 tok/s | 128K | ~5 GB |
Why dan-omni-3b-mobile? Same 3B quality as dan-omni-3b, but 40% smaller file and 40% less RAM. Ideal for phones with tight memory.
Key Differences from dan-omni-3b
| Property | dan-omni-3b | dan-omni-3b-mobile |
|---|---|---|
| File size | 2.0 GB | 1.2 GB |
| Context | 4096 | 2048 |
| Speed | 11.3 tok/s | 9.8 tok/s |
| RAM usage | ~2.5 GB | ~1.5 GB |
| Multimodal | Yes | No |
Dan Omni Model Family
| Model | Size | Speed | Use Case |
|---|---|---|---|
| dan-omni-3b | 3.5 GB | 11.3 tok/s | Full multimodal (text + vision) |
| dan-omni-3b-mobile | 1.2 GB | 9.8 tok/s | Compressed for mobile, 2K context |
| dan-omni-3b-q3s | 1.5 GB | 12.8 tok/s | Aggressive quantization |
| dan-omni-smolm2 | 259 MB | 62.2 tok/s | Ultralight, fastest |
| dan-omni-smolm2-v2 | 259 MB | 62.1 tok/s | Improved quality variant |
Usage
ollama pull sodan/dan-omni-3b-mobile
ollama run sodan/dan-omni-3b-mobile
./llama-cli -m dan-omni-3b-mobile.gguf -p "Hello" --ctx-size 2048
Intended Use
- Phones with ≤3 GB RAM free
- On-device inference with strict memory constraints
- Quick Q&A, translation, short-form tasks
- Offline AI assistant
License
Apache 2.0