AI Models
The three tiers (Palos, Luces, Summum) and their AI pools.
The three tiers
🐔 Palos FREE
Fast and direct. Brain: Mistral Small. For short questions, quick translations, casual chat.
llama-3.1-8b-instant
laguna-xs2
💡 Luces FREE
Balanced. Brain: Llama 3.3 70B. The most-used tier: code, study, medium-depth analysis.
llama-3.3-70b-versatile
llama-4-scout
qwen3-32b
gpt-oss-20b
ling-flash
glm-air
nemotron-nano-9b
nemotron-nano-30b
gemma4-26b
⭐ Summum PREMIUM RECOMMENDED
Max power. Brain: DeepSeek R1 (with GPT-OSS-120B and Llama 3.1 8B as fallbacks). For deep reasoning, complete projects, architecture.
llama-3.3-70b-versatile
llama-4-scout
qwen3-32b
gpt-oss-20b
gpt-oss-120b
kimi-k2
gemini-2.5-flash
deepseek-r1
mistral-large
+ OpenRouter free pool
How to choose
| Use case | Recommended tier |
|---|---|
| Casual conversation, jokes, brief translations | 🐔 Palos |
| Quick code, refactors, simple debugging | 💡 Luces |
| Mid-depth analysis, summaries, study help | 💡 Luces |
| Deep reasoning, math, philosophy | ⭐ Summum |
| Complete projects, software architecture | ⭐ Summum |
| Long-context analysis (huge PDFs) | ⭐ Summum (Gemini Flash, 1M+ tokens) |
| Document/screenshot analysis (vision) | 💡 Luces or ⭐ Summum (Llama 4 Scout) |
Multimodal models (vision)
Some AIs in the pool understand images. If you attach an image to your message, the brain prioritizes one of these:
- Llama 4 Scout — vision + text, 131K context window.
- Gemini 2.5 Flash — vision + text, 1M+ tokens of context.
- Nemotron Nano 12B VL — vision-language compact.
- Nemotron Omni — full multimodal (vision, audio, video).
Tip: You don't usually need to choose the AI manually. The brain does it. The decision you make is the tier (Palos, Luces, Summum), and the brain picks within that tier.
