Best small LLM for the job

Short, data-driven answers to the questions people ask most. Each list is rebuilt from the leaderboard whenever the data changes. Updated November 2026.

Best small open models for coding

1. Qwen3.8 27B · 2. Ternary Bonsai 2 27B · 3. Qwen3.6 27B

Best small open models for math

1. Kolibri-1 · 2. Ternary Bonsai 2 27B · 3. Nanbeige4.3 33B

Best LLMs that run on an 8 GB GPU or laptop

1. Nanbeige4.2 3B · 2. Spark-X2.6 4B · 3. Spark-X2.5 4B

Best LLMs that run on a 16 GB GPU

1. Nanbeige4.2 3B · 2. Spark-X2.6 4B · 3. Spark-X2.5 4B

Best LLMs for a 24 GB GPU (RTX 4090, RTX 3090)

1. Nanbeige4.3 33B · 2. Qwen3.8 27B · 3. Ternary Bonsai 2 27B

Best tiny LLMs for phones and edge devices

1. Nanbeige4.2 3B · 2. Spark-X2.6 4B · 3. Spark-X2.5 4B

Best cheap LLM APIs (open models)

1. Qwen3.8 27B · 2. Qwen3.6 35B-A3B · 3. Gemma 4 31B

Best small open vision-language models

1. Qwen3.8 27B · 2. Ternary Bonsai 2 27B · 3. Qwen3.6 27B

Best open models with an Apache 2.0 or MIT license

1. Nanbeige4.3 33B · 2. Qwen3.8 27B · 3. Ternary Bonsai 2 27B

Best European open language models

1. Kolibri-1 · 2. Ministral 3 14B · 3. Magistral Small 1.2

Best American open models under 100B

1. Ternary Bonsai 2 27B · 2. Nemotron 3.5 Bolt 74B-A9B · 3. Muse Glimmer 30B

Best fully open models (open data and code)

1. K2-V2 Instruct · 2. OLMo 3.1 32B Think · 3. OLMo 3 7B Think