Supported Models

Gopher supports multiple LLM providers and models for strategy generation and backtesting.

Providers

Provider	Description	API Key Required
Gopher Credits	Hosted inference, easiest setup	Yes (Gopher Key)
OpenRouter	Access 100+ models with one key	Yes
OpenAI	GPT-4, GPT-4o models	Yes
Ollama	Local models	No
Basilica	OpenAI-compatible deployments	No (deployment URL)
Custom	Any OpenAI-compatible API	Varies

Gopher Credits (Recommended)

The easiest way to use Gopher - no external API accounts needed.

Setup

1. Go to gotrader.gopher-ai.com/settings
2. Create an account or sign in
3. Purchase credits
4. Copy your Gopher Key (format: gopher_xxx...)
5. Enter in Gopher Settings or set as environment variable:

   export BART_GOPHER_CODE='gopher_your-key-here'

Benefits

• No external accounts: Single account for everything
• Optimized models: Pre-configured for best results
• Pay-as-you-go: Only pay for what you use
• Simple setup: Just one key to configure

When using Gopher Credits, model selection is handled automatically with optimized defaults.

Recommended Models

For Strategy Generation (Loop Model)

These models are best for the main evolution loop - generating and refining strategies.

Model	Provider	Quality	Speed	Cost
qwen/qwen3-max	OpenRouter	Excellent	Medium	$$
anthropic/claude-3.5-sonnet	OpenRouter	Excellent	Fast	$$$
openai/gpt-4-turbo	OpenRouter/OpenAI	Excellent	Medium	$$$
deepseek/deepseek-chat	OpenRouter	Very Good	Fast	$
meta-llama/llama-3.1-70b-instruct	OpenRouter	Good	Medium	$

Recommendation: qwen/qwen3-max offers the best balance of quality and cost.

For Backtesting (Backtest Model)

These models make trade decisions during backtests. Speed and efficiency matter here.

Model	Provider	Quality	Speed	Cost
qwen/qwen3-vl-8b-instruct	OpenRouter	Good	Very Fast	$
openai/gpt-4o-mini	OpenRouter/OpenAI	Good	Fast	$
meta-llama/llama-3.1-8b-instruct	OpenRouter	Good	Very Fast	$
mistral/mistral-7b-instruct	OpenRouter	Good	Very Fast	$

Recommendation: qwen/qwen3-vl-8b-instruct is fast and cost-effective.

OpenRouter Models

OpenRouter provides access to models from multiple providers with a single API key.

Getting Started

1. Create an account at openrouter.ai
2. Generate an API key at openrouter.ai/keys
3. Add credits to your account
4. Enter the key in Gopher Settings

Popular OpenRouter Models

Premium Tier:

• anthropic/claude-3.5-sonnet - Fast, excellent reasoning
• anthropic/claude-3-opus - Most capable Claude model
• openai/gpt-4-turbo - Latest GPT-4
• openai/gpt-4o - Optimized GPT-4

Mid Tier:

• qwen/qwen3-max - Excellent quality/cost ratio
• deepseek/deepseek-chat - Good for technical tasks
• meta-llama/llama-3.1-70b-instruct - Strong open model

Budget Tier:

• qwen/qwen3-vl-8b-instruct - Fast and efficient
• openai/gpt-4o-mini - Affordable GPT-4 variant
• meta-llama/llama-3.1-8b-instruct - Good performance/cost

OpenAI Models

Use OpenAI directly with an OpenAI API key.

Available Models

Model ID	Description
gpt-4-turbo	Latest GPT-4, 128k context
gpt-4o	Optimized GPT-4, faster
gpt-4o-mini	Smaller, more affordable
gpt-4	Original GPT-4
gpt-3.5-turbo	Fast, budget option

Getting Started

1. Create account at platform.openai.com
2. Generate API key at platform.openai.com/api-keys
3. Add credits to your account
4. Enter the key in Gopher Settings

Local Models (Ollama)

Run models locally without API costs using Ollama.

Setup

1. Install Ollama from ollama.ai
2. Pull a model:

   ollama pull llama3.1:8b

3. Start Ollama (usually runs automatically)
4. In Gopher Settings, add a custom model:
   • Model ID: llama3.1:8b
   • Provider: Ollama
   • Base URL: http://localhost:11434/v1
   • API Key: (leave empty)

Recommended Local Models

Model	Size	VRAM Required
llama3.1:8b	4.7 GB	8 GB
llama3.1:70b	40 GB	48 GB
mistral:7b	4.1 GB	8 GB
codellama:13b	7.4 GB	16 GB
qwen2:7b	4.4 GB	8 GB

Note: Local models may produce lower quality results than cloud models, especially for smaller sizes.

Custom Models

Add any OpenAI-compatible API endpoint.

Adding a Custom Model

1. Go to Settings > Models
2. Click + Add Custom
3. Fill in:
   • Model ID: The model identifier
   • Display Name: Friendly name
   • Provider: Select "Custom"
   • Base URL: OpenAI-compatible root URL only (e.g., https://api.example.com/v1)
   • API Key: Your key for this endpoint

Note: Do not include /chat/completions in the base URL. For per-loop overrides, see Configuration → Inference.

Compatible Services

Many services offer OpenAI-compatible APIs:

• Together AI: together.ai
• Anyscale: anyscale.com
• Perplexity: perplexity.ai
• Groq: groq.com
• Basilica: docs.basilica.ai
• Local LLMs: LM Studio, vLLM, text-generation-webui

Model Selection Tips

Quality vs Cost

Priority	Loop Model	Backtest Model
Best Quality	claude-3.5-sonnet	gpt-4o-mini
Balanced	qwen3-max	qwen3-vl-8b
Budget	deepseek-chat	llama-3.1-8b
Free (Local)	llama3.1:70b	llama3.1:8b

Cost Optimization

1. Use a capable model for the Loop (strategy generation)
2. Use a fast/cheap model for Backtest (trade decisions)
3. Reduce iterations if costs are too high
4. Monitor usage in your provider dashboard

Testing Models

Click Test next to any model in Settings to verify:

• API key is valid
• Endpoint is reachable
• Model responds correctly

Fallback Model

Configure a fallback model in Settings for resilience:

• Used when primary model fails
• Should be a reliable, well-supported model
• Default: anthropic/claude-sonnet-4.5

Next Steps

• Configuration - Set up models and API keys
• Quick Start - Run your first evolution
• Troubleshooting - API key issues