Models & Aliases

llmist supports multiple LLM providers with convenient aliases for quick access.

Model Aliases

Use short aliases instead of full model names:

Alias	Full Model Name	Provider
`gpt5`	`gpt-5`	OpenAI
`gpt5-mini`	`gpt-5-mini`	OpenAI
`gpt4o`	`gpt-4o`	OpenAI
`o4-mini`	`o4-mini`	OpenAI
`sonnet`	`claude-sonnet-4-5`	Anthropic
`opus`	`claude-opus-4-5`	Anthropic
`haiku`	`claude-haiku-4-5`	Anthropic
`flash`	`gemini-2.5-flash`	Google
`pro`	`gemini-3-pro-preview`	Google

Provider-Prefixed Models

For explicit provider selection, use the provider:model format:

    # Explicit provider selection
    npx @llmist/cli complete "Hello" --model openai:gpt-5
    npx @llmist/cli complete "Hello" --model anthropic:claude-sonnet-4-5
    npx @llmist/cli complete "Hello" --model gemini:gemini-2.5-flash
    npx @llmist/cli complete "Hello" --model huggingface:meta-llama/Llama-3.1-8B-Instruct

Model Capabilities

Model	Vision	Streaming	Tool Use	Reasoning	Context
GPT-5	✓	✓	✓	✓	128K
GPT-5 Mini	✓	✓	✓	✗	128K
GPT-4o	✓	✓	✓	✗	128K
o3	✗	✓	✓	✓	200K
o4-mini	✗	✓	✓	✓	200K
Claude Opus 4.5	✓	✓	✓	✓	200K
Claude Sonnet 4.5	✓	✓	✓	✓	200K
Claude Haiku 4.5	✓	✓	✓	✓	200K
Gemini Flash	✓	✓	✓	✓	1M
Gemini Pro	✓	✓	✓	✓	1M
DeepSeek V3.2	✗	✓	✓	✗	64K
DeepSeek R1	✗	✓	✓	✗	64K
Llama 3.3 70B	✗	✓	✗	✗	128K
Qwen 2.5 72B	✗	✓	✗	✗	128K
Mixtral 8x7B	✗	✓	✗	✗	32K

Recommended Models by Use Case

Use Case	Recommended	Why
General tasks	`sonnet`	Best balance of quality and speed
Complex reasoning	`opus`, `o3`, DeepSeek R1	Highest capability; use `.withReasoning("high")` for o3/Claude
High-volume tasks	`haiku`, `flash`	Fast and cost-effective
Long documents	`flash`, `pro`	1M token context
Coding	`sonnet`, `gpt5`, DeepSeek V3.2	Strong code understanding
Vision tasks	`gpt4o`, `flash`	Excellent image analysis
Open-source/self-hosted	Llama 3.3 70B, Qwen 2.5 72B	Free via HuggingFace

HuggingFace Models

llmist supports popular open-source models via HuggingFace’s serverless inference:

DeepSeek Family

deepseek-ai/DeepSeek-V3.2 - 685B MoE model for general reasoning and tool use
deepseek-ai/DeepSeek-R1 - Reasoning model excelling at math, logic, and coding
deepseek-ai/DeepSeek-Coder-V2-Instruct - Specialized for code generation

Meta Llama

meta-llama/Llama-3.3-70B-Instruct - General-purpose flagship model
meta-llama/Llama-3.1-8B-Instruct - Efficient smaller variant
meta-llama/Llama-3.2-11B-Vision-Instruct - Vision-enabled model

Qwen (Alibaba)

Qwen/Qwen2.5-72B-Instruct - Strong general-purpose model
Qwen/Qwen2.5-Coder-32B-Instruct - Code-specialized
Qwen/Qwen2-VL-72B-Instruct - Vision-language model

Mistral AI

mistralai/Mixtral-8x7B-Instruct-v0.1 - Mixture-of-experts architecture
mistralai/Mistral-Nemo-Instruct-2407 - 12B efficient model

Image Generation Models

Model	Provider	Description
`dall-e-3`	OpenAI	High-quality image generation
`dall-e-2`	OpenAI	Faster, lower cost
`imagen-3`	Google	Gemini image generation

Speech Models

Model	Provider	Description
`tts-1`	OpenAI	Text-to-speech, standard quality
`tts-1-hd`	OpenAI	Text-to-speech, high quality

import { LLMist } from 'llmist';

// Using alias
const answer = await LLMist.createAgent()
  .withModel('sonnet')
  .askAndCollect('Hello!');

// Using full name
const answer2 = await LLMist.createAgent()
  .withModel('anthropic:claude-sonnet-4-5')
  .askAndCollect('Hello!');

// HuggingFace model
const answer3 = await LLMist.createAgent()
  .withModel('huggingface:meta-llama/Llama-3.1-8B-Instruct')
  .askAndCollect('Hello!');

// HuggingFace with provider routing
const answer4 = await LLMist.createAgent()
  .withModel('hf:deepseek-ai/DeepSeek-V3.2:fastest')
  .askAndCollect('Write a Python function to calculate factorial');

// OpenRouter model (access 400+ models)
const answer5 = await LLMist.createAgent()
  .withModel('openrouter:deepseek/deepseek-chat')
  .askAndCollect('Explain quantum computing');

// OpenRouter with alias
const answer6 = await LLMist.createAgent()
  .withModel('or:sonnet')
  .askAndCollect('Hello!');

# Using alias
npx @llmist/cli complete "Hello" --model sonnet

# Using full name
npx @llmist/cli complete "Hello" --model anthropic:claude-sonnet-4-5

# HuggingFace model
npx @llmist/cli complete "Hello" --model huggingface:meta-llama/Llama-3.1-8B-Instruct

# HuggingFace with provider routing (use fastest endpoint)
npx @llmist/cli complete "Hello" --model hf:Qwen/Qwen2.5-72B-Instruct:fastest

# OpenRouter (access 400+ models via unified gateway)
npx @llmist/cli complete "Hello" --model openrouter:deepseek/deepseek-chat

# OpenRouter with alias
npx @llmist/cli agent "Review my code" --model or:sonnet

Auto-Discovery

llmist automatically discovers available providers based on environment variables:

Variable	Provider
`OPENAI_API_KEY`	OpenAI
`ANTHROPIC_API_KEY`	Anthropic
`GEMINI_API_KEY`	Google Gemini
`OPENROUTER_API_KEY`	OpenRouter
`HF_TOKEN`	HuggingFace

See Environment Variables for complete configuration.