This is a snapshot of the popular LLMs that are available on the Hatz AI platform.
DISCLAIMER: The newest models from each developer tend to be the "best" option. Many LLMs return similar or comparable results for the majority of business use cases.
Don't worry about choosing the "right" model - try several of the newest ones and choose your favorite.
Newest models from each developer are marked in blue.
Cheap = lower credit consumption
Developer | Model | What Is It Good At? |
Amazon | Nova Micro | fast text tasks, lower credit usage |
Amazon | Nova Lite | fast summaries + chat |
Amazon | Nova Pro | reasoning + coding |
Amazon | Nova Lite 2 | cheaper, fast general chat |
Anthropic | Claude 3 Haiku | very fast summaries |
Anthropic | Claude 3 Sonnet | writing + analysis |
Anthropic | Claude 3.5 Sonnet v2 | writing + coding |
Anthropic | Claude 3.7 Sonnet | writing + coding |
Anthropic | Claude 3.5 Haiku | fast responses |
Anthropic | Claude 4 Sonnet | agents + coding |
Anthropic | Claude 4 Opus | reasoning + coding |
Anthropic | Claude 4.5 Opus | reasoning + agents |
Anthropic | Claude 4.5 Sonnet | strong all‑purpose |
Anthropic | Claude 4.5 Haiku | cheapest fast agents |
DeepSeek | DeepSeek R1 | hard problem solving |
Gemini 1.5 Flash | fast long-doc summaries | |
Gemini 2.0 Flash | fast general chat | |
Gemini 2.0 Flash Lite | cheapest bulk text | |
Gemini 1.5 Pro | deep long-doc analysis | |
Gemini 2.5 Flash | fast production work | |
Gemini 2.5 Flash Image (Nano Banana) | fast image generation | |
Gemini 2.5 Pro | best Gemini reasoning | |
Gemini 3 Pro Preview | newest Gemini (testing) | |
Gemini 3 Pro Image (Nano Banana Pro) | best image generation | |
Meta | Llama 3.1 70B Instruct | quality open chat |
Meta | Llama 3.1 8B Instruct | cheap open chat |
Meta | Llama 3.2 1B Instruct | ultra-cheap, tiny tasks |
Meta | Llama 3.2 11B Instruct | balanced open chat |
Meta | Llama 4 Maverick 17B Instruct | stronger mid-size chat |
Meta | Llama 4 Scout 17B Instruct | fast drafting + agents |
Mistral AI | Mistral 8x7B Instruct | efficient bulk summaries |
MistralAI | Mistral Large | high-quality writing |
MistralAI | Mistral Large 3 | best Mistral reasoning |
MistralAI | Mistral 7B Instruct | cheap basic chat |
OpenAI | GPT-4.1 | coding + analysis |
OpenAI | GPT-4.1 Mini | cheap general assistant |
OpenAI | GPT-4.1 Nano | ultra-cheap bulk text |
OpenAI | GPT 4.5 Preview | new (testing) |
OpenAI | GPT 4o | vision + chat |
OpenAI | GPT 3.5 Turbo | cheap basic chat |
OpenAI | o1 | reasoning |
OpenAI | o3-mini | cheaper reasoning |
OpenAI | GPT 4 | strong general |
OpenAI | o3 | strong reasoning |
OpenAI | o4-mini | fast cheap reasoning |
OpenAI | GPT-5 nano | fastest, cheap text |
OpenAI | GPT-5 mini | cheap agents + summaries |
OpenAI | GPT-5 Chat | chat assistant |
OpenAI | GPT-5 | strong general + coding |
OpenAI | GPT-5.1 | strong quality general |
OpenAI | GPT-5.2 | top reasoning + coding |
xAI | Grok 2 | general chat |
xAI | Grok 2 Vision | vision + OCR |
xAI | Grok 3 | reasoning + agents |
xAI | Grok 3 Fast | fastest chat |
xAI | Grok 3 Mini | cheap general |
xAI | Grok 3 Mini Fast | cheap + fast |
xAI | Grok 4 | research + reasoning |
Some LLM Terminology:
Token
The smallest unit of data that an LLM processes
Usually about 3/4 of a word
If you want to see how many INPUT tokens something is, use this site (it works best for OpenAI models):
***In some cases, tokens are not just words. If you upload a PDF with images, tokens will applied to interpret the images in addition to the words. If you upload a PNG or JPG with no words, tokens will still be used to understand the image.
Context Window
The segment of text that an AI model considers at any given moment when processing or generating language
You can think of this as the size of the input
If you upload a file (PDF, PNG, XLSX, etc) that is below the 30MB limit, but it is larger than the context window for the selected LLM, the request will fail. Try switching to a larger model, like Gemini
Knowledge Cutoff Date
The date that the LLM stopped "learning" new information. If you ask GPT 3.5 Turbo about an event that happened yesterday, it will not have that data
Reasoning
Reasoning models are a type of AI model that are trained to engage in "thinking" before providing an answer, breaking down problems into steps and showing their reasoning process, unlike models that directly generate answers.
They can take more time to answer because they are reflecting on their analysis step-by-step.
Vision
Computer vision gives LLMs the ability to "see" and interpret visual data like images and videos, enabling them to identify objects, classify scenes, and perform tasks that mimic human vision.
More Info
To dive deeper into these LLMs, check out these resources*
*Some features outlined on these sites might not be available on the Hatz AI platform, or they might work differently than explained here. Use these articles to better understand the token limits, reasoning models, and use cases.