LLM Hallucination Index RAG SpecialModel Insights
qwen1.5-32b-chat
Details
Developer
Alibaba
License
Tongyi Qianwen
Model parameters
32b
Supported context length
32k
Price for prompt token
$0.8/Million tokens
Price for response token
$0.8/Million tokens
Model Performance Across Task-Types
Chainpoll Score
Short Context
0.87
Medium Context
0.99
Model Insights Across Task-Types
Digging deeper, here’s a look how qwen1.5-32b-chat performed across specific datasets
Short Context RAG
Medium Context RAG
This heatmap indicates the model's success in recalling information at different locations in the context. Green signifies success, while red indicates failure.
Performance Summary
| Tasks | Task insight | Cost insight | Dataset | Context adherence | Avg response length |
|---|---|---|---|---|---|
| Short context RAG | The model demonstrates good reasoning and comprehension skills, excelling at short context RAG. It also decent mathematical proficiency, as evidenced by its performance on DROP and ConvFinQA benchmarks. It could have performed better for its size. | It is costly model for the performance it offers. We recomend using 8x cheaper Gemini Flash or 3x cheaper Haiku. You can also go ahead with 4x cheaper Llama-3-8b which offers a better performance at low cost. | Drop | 0.87 | 283 |
| Hotpot | 0.86 | 283 | |||
| MS Marco | 0.89 | 283 | |||
| ConvFinQA | 0.87 | 283 | |||
| Medium context RAG | Flawless performance making it suitable for any context length upto 20000 tokens. It struggles a bit for 25000 but can be used without much issues. | Great performance but we recommed using 8x cheaper Gemini Flash. | Medium context RAG | 0.99 | 283 |