bartowski/Llama-3.2-3B-Instruct-GGUF:Q4_K_M
bartowski/Llama-3.2-1B-Instruct-GGUF:Q4_K_M
bartowski/Qwen2.5-Coder-3B-Instruct-GGUF:Q4_K_M
bartowski/Qwen2.5-Coder-1.5B-Instruct-GGUF:Q4_K_M
bartowski/Phi-3.5-mini-instruct-GGUF:Q4_K_M
bartowski/DeepSeek-R1-Distill-Qwen-1.5B-GGUF:Q4_K_M
mradermacher/SmolLM2-1.7B-Instruct-GGUF:Q4_K_M
unsloth/Qwen2.5-Omni-3B-GGUF
unsloth/Qwen3.5-0.8B-MTP-GGUF
Abiray/LFM2.5-2.6B-Heretic-Abliterated-GGUF
Abiray/LFM2.5-2.6B-Heretic-Abliterated-GGUF
QuantFactory/VersatiLlama-Llama-3.2-3B-Instruct-Abliterated-GGUF
llama serve -hf
OpenWebUI + OpenRouter Optimization Reference Guide
This guide details configuration settings, model routing strategies, custom headers, and system prompts designed to maximize API balance longevity on OpenRouter when connecting via OpenWebUI.
1. Top Recommended Models on OpenRouter
Use these exact model IDs in OpenWebUI depending on task complexity:
| Task Tier | OpenRouter Model ID | Approximate Cost (Input / Output) | Primary Use Case |
|---|---|---|---|
| Tier 0: Free | nvidia/llama-3.1-nemotron-70b-instruct:free |
$0.00 | Simple questions, copy editing, text formatting, basic scripts. |
| Tier 1: Budget Workhorse | deepseek/deepseek-chat |
~$0.26 / $1.00 per 1M tokens | Daily programming, logic tasks, general writing, technical Q&A. |
| Tier 1: High Context & Speed | google/gemini-2.5-flash |
~$0.15 / $1.25 per 1M tokens | Large document processing, long transcripts, fast reference searches. |
| Tier 2: Frontier / Heavy Duty | anthropic/claude-3.5-sonnet |
~$3.00 / $15.00 per 1M tokens | Complex architectural refactoring, multi-step edge-case debugging. |
Tip: You can set default access permissions and visible model lists in OpenWebUI via Admin Panel → Settings → Models.
2. OpenWebUI System Parameters (Context Controls)
To prevent conversation threads from sending massive token volumes back to OpenRouter on every turn, configure these parameters under User Settings → System / Advanced (or globally in Admin Panel):
# OpenWebUI System Parameters Configuration
Context Length (Num Ctx): 8192
# Limits maximum window context sent per request to cap token costs per turn.
History Offsets / Chat Context: 10 - 15 messages
# Prevents sending 50+ previous conversation turns back to the API on follow-up questions.
System Prompt Override: Active
# Uses a concise system prompt to minimize system-level token consumption.