Skip to content

Free AI Models Guide

AI Models

Last Updated: April 2026

Table of Contents

  1. Hugging Face Inference Providers
  2. Other Free Tiers
  3. Recommendations by Use Case
  4. Model Comparison Tables

Free AI Models Overview

Hugging Face Inference Providers

OpenCode natively supports Hugging Face Inference Providers - giving you access to open models from 17+ providers.

Quick Setup

Hugging Face Setup

# 1. Create token at huggingface.co/settings/tokens
#    (needs "Make calls to Inference Providers" permission)

# 2. Run auth login
opencode auth login

# 3. Select Hugging Face when prompted
# Enter your token: hf_

# 4. Select a model
/models

Best Coding Models

Model Best For Provider Context Notes
Qwen2.5-Coder-32B Code reasoning Featherless 131K Good for complex code tasks

Other Notable Models

For Reasoning:

  • DeepSeek-R1 - Chain-of-thought reasoning, great for complex logic
  • Kimi-K2.5 - Fast, good for general tasks
  • GLM-4.7 - Free tier available

For General Use:

  • Llama 3.1 8B - Fast, cheapest
  • Qwen2.5 7B - Good balance
  • Gemma 4 31B - Google's best

Resources:

OpenCode Free Models

OpenCode offers built-in free models through OpenCode Zen - a curated set of models developed by Zen Labs that have been tested and benchmarked specifically for coding agents.

Getting Started

OpenCode comes with free models ready to use. Just run: 

opencode
/models
# Select from available free models

Available Free Models

Model Context Best For Notes
Big Pickle 200K Complex coding Stealth model, optimized for agents
MiniMax M2.5 Free 128K+ General coding Strong at tool use
Qwen3.6 Plus Free 131K+ Complex tasks High performance, reasoning capable
MiMo V2 Pro Free 131K+ Fast inference One of the fastest available
Nemotron 3 Super Free 1M Long context NVIDIA's open-weight agent model

Limited Time

These free models are available while OpenCode collects feedback. During this period, data may be used to improve the model. For sensitive code, consider paid options or local models (Ollama).

About Big Pickle

Big Pickle is a stealth model developed by Zen Labs, the team behind OpenCode. It's optimized specifically for coding agents with a 200K token context window - the largest among free options.

How to Access

opencode

# In terminal:
/models

# Select from available free models

Other Free Tiers

Free Tier Comparison

Provider Free Credits Best Models Sign Up
Google AI Studio 15 RPM, 250K TPM Gemini 2.5 Pro/Flash aistudio.google.com
GitHub Models 50-150 req/day o3-mini, GPT-4.1 github.com/models
NVIDIA NIM 1,000 credits DeepSeek R1, Llama build.nvidia.com
Hugging Face Monthly credits 300+ models huggingface.co
Groq Low-cost Llama 3.3, DeepSeek R1, Qwen QwQ console.groq.com
xAI $25 credits Grok 4 x.ai

Best Free Option

  • Gemini 2.5 Pro (best free model)
  • Gemini 2.5 Flash (fastest)
  • 1M token context
  • Sign up at aistudio.google.com

GitHub Models

  • GPT-4.1 (excellent coding)
  • o3-mini (reasoning)
  • Direct integration with OpenCode via /connect

Resources:

Groq (Fastest Inference)

Fastest Inference Available

Groq is not free - but it offers the fastest inference speed available (~1000+ tokens/second). It's low-cost and integrates natively with OpenCode.

Model Context Notes
llama-3.3-70b-versatile 128K Best overall
deepseek-r1-distill-llama-70b 128K Strong reasoning
qwen-qwq-32b 128K Fast reasoning
gemma-2-9b-it 8K Lightweight

Setup

  1. Get a free API key at console.groq.com/keys
  2. In OpenCode, run /connect → search for "Groq" → paste your API key
  3. Run /models to select a Groq model

Recommendations by Use Case

Best for Coding Tasks

  1. Qwen2.5-Coder-32B (HF) - Code reasoning
  2. GPT-4.1 (GitHub Models) - General coding
  3. Gemini 2.5 Pro (Google AI Studio) - Long context

Best for Reasoning

  1. DeepSeek-R1 (HF Hyperbolic) - Chain-of-thought
  2. o3-mini (GitHub Models) - Reasoning
  3. Gemini 2.5 Pro (Google) - Long context

Best for Speed

  1. Groq - Fastest inference (1000+ t/s)
  2. Qwen3-Coder-Next - 128 t/s

Best for Free Tier (No Credit Card)

  1. Big Pickle (OpenCode built-in)
  2. GLM-4.7 Flash (HF)
  3. Gemini 2.5 Flash (Google)

Model Comparison Tables

Coding Models Comparison

Model Provider Context Speed Notes Best For
Qwen2.5-Coder-32B Featherless 131K Medium Good code reasoning Code reasoning
DeepSeek-R1 Hyperbolic 131K Medium Chain-of-thought Complex reasoning
GPT-4.1 GitHub 32K Fast Free tier General coding

Free Models Comparison

Model Source Notes
Big Pickle OpenCode Works out of box
GLM 4.7 Flash Hugging Face Slower
Gemini 2.5 Flash Google Generous free tier
Gemma 4 31B Hugging Face Google's best open model

TODO: Review Schedule

  • Review monthly - check for new models
  • Check free tier limits changed
  • Update recommendations based on benchmarks

Resources

This is a living document. Revisit and update regularly.