Shortlist
What is the best open source LLM in 2026?
Kimi K3 is our pick, scoring 87.5 out of 100 across craft, speed, control and value, with GLM-5.2 behind it on 87.1. These are the five best open-weights models you can host yourself, chosen from 19 that qualify and ranked on the same rubric as the rest of the index.
Index updated
Chosen from 19 that qualify, out of 53 tracked
The shortlist
- 01
Kimi K3
The largest open-weights model anyone has shipped, and it reads a million tokens without blinking.
Open weightsFree tierBest for
Long-document and agent work at the top of the open field
- 02
GLM-5.2
Frontier-class output, an MIT licence, and a million tokens of context. That combination is the story.
Open weightsFree tierBest for
Frontier-class work you need to keep in house
- 03
Qwen 3.6
The open Qwen release most people should actually run, in sizes that fit real hardware.
Open weightsFree tierBest for
Self-hosting and fine-tuning on hardware you already have
- 04
Mistral Large 3
Europe's answer, published under Apache 2.0, with a quarter of a million tokens of context.
Open weightsFree tierBest for
European teams that need capability and data residency at once
- 05
DeepSeek V4 Flash
Near-frontier answers at roughly a hundredth of frontier prices, with the weights published.
Open weightsFree tierBest for
High-volume work where the token bill is the constraint
