Free llm rotation

Skill TheWatcher01/skills/.claude/skills/free-llm-rotation

Configure and use free LLM providers with intelligent rotation and fallback. USE when: setting up ZeroClaw with free tiers, hitting rate limits, switching providers dynamically, comparing free model quality, configuring Ollama local fallback, or when the user says "use a free model", "I hit the rate limit", "switch to a cheaper model", "configure Ollama". Covers OpenRouter (26+ free), Groq (free tier), Ollama (local), Cerebras.From its SKILL.md

Install

npx -y skills add TheWatcher01/skills --skill free-llm-rotation

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.7 KB, ~1.4k tokens by cl100k_base, as published. Nobody here has run it

Skill: Free LLM Provider Rotation

Free Provider Landscape (March 2026)

Provider	Free Models	Speed	Rate Limit	Best For
OpenRouter	26+ free models	Varies	Per-model	Variety, fallback
Groq	llama-3.1-8b, llama-3.3-70b	~560-280 tps	1K RPM	Speed, daily use
Ollama	Any local model	CPU: ~10 tps	None	Always-on fallback
Cerebras	Llama 3.3 70B	Very fast	Free tier	Quick tasks

Top Free Models on OpenRouter (March 2026)

Model	Context	Speed	Best For
`openai/gpt-oss-120b:free`	131K	~500 tps	General, reasoning
`meta-llama/llama-4-scout-17b-16e-instruct:free`	131K	~750 tps	Fast tasks
`nvidia/nemotron-3-nano-30b-a3b:free`	256K	Fast	Agent tasks
`arcee-ai/trinity-large-preview:free`	128K	Medium	Creative, agent
`openai/gpt-oss-20b:free`	131K	~1000 tps	Ultra-fast

ZeroClaw Free Tier Configuration

# ~/.zeroclaw/config.toml

# Primary: OpenRouter (free tier)
# Set ZEROCLAW_API_KEY=<openrouter-key>
default_provider = "openrouter"
default_model    = "openai/gpt-oss-120b:free"

# Fallback chain: Groq → Ollama
[reliability]
fallback_providers = ["groq", "ollama"]
provider_retries   = 2

# Groq provider (set GROQ_API_KEY env var)
[model_providers.groq]
name     = "groq"
base_url = "https://api.groq.com/openai/v1"

# Ollama local (no API key, always free)
[model_providers.ollama]
name     = "ollama"
base_url = "http://127.0.0.1:11434"

# Model routes for manual control
[[model_routes]]
hint     = "fast"
provider = "groq"
model    = "llama-3.1-8b-instant"

[[model_routes]]
hint     = "local"
provider = "ollama"
model    = "llama3.2:3b"

[[model_routes]]
hint     = "smart"
provider = "openrouter"
model    = "openai/gpt-oss-120b:free"

How to Get Free API Keys

OpenRouter (primary — 26+ free models)

Go to https://openrouter.ai/
Sign up (GitHub login OK)
API Keys → Create key
Free models: filter by $0/M tokens
Usage: generous free tier, no credit card

Groq (ultra-fast, free tier)

Go to https://console.groq.com/
Sign up → API Keys → Create
Free: 1K RPM, 30K RPD for most models
Set GROQ_API_KEY=gsk_...

Cerebras (optional, very fast)

https://cloud.cerebras.ai/ → Sign up
Free tier available
Set CEREBRAS_API_KEY=...

Ollama Management (local, always free)

# List installed models
ollama list

# Pull models (GPU-first, CPU fallback configured)
ollama pull llama3.2:3b      # 2GB — fast, good for general tasks
ollama pull qwen2.5-coder:7b # 4.7GB — best free code model
ollama pull nomic-embed-text # 274MB — embeddings (free!)

# Test GPU or CPU usage
ollama run llama3.2:3b "test" 2>&1 | head -5
journalctl -u ollama | grep "inference compute"
# GPU: "id=gpu0 library=cuda ..."
# CPU: "id=cpu library=cpu ..."

# Ollama API
curl http://127.0.0.1:11434/api/generate -d '{
  "model": "llama3.2:3b",
  "prompt": "Hello",
  "stream": false
}'

Rate Limit Handling

When you hit a rate limit (429 error):

ZeroClaw auto-falls to next provider in fallback_providers
If all cloud providers rate-limited → Ollama (local, no limit)

Manual routing:

# Force local model
zeroclaw agent -m "hint:local explain this code: ..."

# Force fast model  
zeroclaw agent -m "hint:fast summarize this document: ..."

Inject API Keys via Infisical

# Store keys in Infisical (self-hosted)
infisical secrets set ZEROCLAW_API_KEY="or-..." --env=dev
infisical secrets set GROQ_API_KEY="gsk_..." --env=dev

# Run daemon with injected secrets
infisical run --env=dev -- zeroclaw daemon

# Verify
infisical secrets list --env=dev

Model Selection Guide

Task type → Recommended model (free)

Code generation/review → openai/gpt-oss-120b:free OR groq:llama-3.1-8b-instant
Quick Q&A / routing → groq:llama-3.1-8b-instant (fastest)
Long document analysis → openai/gpt-oss-120b:free (131K context)
Always-available / offline → ollama:llama3.2:3b
Embeddings (RAG) → ollama:nomic-embed-text (free, local)

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.