A comparison of various language models under 9GB in size highlights their performance across different benchmarks. The data, sourced from lab reports, includes details on parameters, context length, modality, and licensing for models such as Qwen3.5-9B, Gemma 4 E4B, and LFM2.5-8B-A1B. Performance metrics for math and coding tasks are presented, with specific scores for each model on benchmarks like AIME, HMMT, MATH-500, and SWE-bench. AI
IMPACT Provides comparative data for developers choosing smaller, efficient AI models for specific tasks.
RANK_REASON The item compares performance metrics of various AI models on specific benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]
- Gemma
- Gemma 4 E4B
- Granite 4.2-8B
- IBM
- K2-Horizon-7B
- LFM2.5-8B-A1B
- Liquid
- Llama 3
- MiMo-V2.6-Distill-Qwen-9B
- Mistral AI
- Nous Hermes
- OpenHermes
- Orca
- Phi 3
- Qwen
- Qwen3.5-9B
- Starling
- Xiaomi
- Zephyr
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →