Best small model
The best small LLM runs on your own hardware without giving up too much intelligence. These are the strongest open-weights models between 4B and 40B parameters we track.
#1 for Best small modelAlibaba Qwen
Qwen3.8 27B
Qwen3.8 27B is the strongest compact model we track — the highest Artificial Analysis Intelligence Index among open-weights models between 4B and 40B parameters that you can self-host.
See the full breakdown1
Qwen3.8 27B
Alibaba Qwen
Input
$0.50
Output
$3.00
Context
262K
58.6
2K2 Horizon MoVA 36B A4B
Institute of Foundation Models
Input
—
Output
—
Context
524K
43.1
3K2 Horizon 7B
Institute of Foundation Models
Input
—
Output
—
Context
524K
36.2
4Qwen3.6 35B A3B
Alibaba Qwen
Input
$0.38
Output
$2.25
Context
262K
31.0
5Gemma 4 26B A4B
Google DeepMind
Input
$0.10
Output
$0.37
Context
262K
29.3
6Muse Glimmer
Meta
Input
$0.32
Output
$1.35
Context
131K
29.3
7Granite 4.2 30B
IBM
Input
$0.16
Output
$0.65
Context
131K
25.9
Get the weekly LLM rankings
One email a week: what moved, what's new, and which model to actually use for your work. No hype, no benchmark jargon.