Below is an overview of each LLM model that was used in the project.
| Model ID | Provider | Model Type | Parameters | Highlights |
|---|---|---|---|---|
anthropic/claude-3.5-sonnet |
Anthropic | Instructional LLM | ~Unlimited | Fast, accurate Claude family model |
google/gemini-2.0-flash-001 |
Google DeepMind | Multimodal LLM | - | Lightweight version of Gemini 2 |
openai/gpt-4o-2024-11-20 |
OpenAI | Multimodal LLM | - | Fastest GPT-4 variant, vision-capable |
openai/o1 |
OpenAI | Reasoning LLM | - | Previous full o-series reasoning model |
openai/o1-mini |
OpenAI | Lightweight LLM | - | Cost-effective, fast inference |
openai/o3-mini |
OpenAI | Lightweight LLM | - | Smaller sibling in the OpenAI stack |
deepseek/deepseek-chat |
DeepSeek | Chat Model | 7B | Optimized for dialogue tasks |
deepseek/deepseek-chat-v3-0324 |
DeepSeek | Chat Model | 67B | Latest flagship model (Mar 2024) |
deepseek/deepseek-r1 |
DeepSeek | Reasoning LLM | 236B tokens | Trained on custom long-context corpus |
qwen/qwq-32b |
Alibaba/Qwen | Instruction LLM | 32B | High-performance multilingual model |
meta-llama/llama-3.3-70b-instruct |
Meta | Instruction LLM | 70B | Open-source instruction-tuned LLaMA 3 |