Open-weight model guide
An evidence-led directory of the open-weight models in benchr's current index, with license, model ID, context, pricing when published, and deployment routes.
Open-weight model guide
Open weights do not mean zero operating cost or identical licenses. The directory keeps license and hosted API fields visible and links to local-compute and cost guidance.
Browse models
These cards are derived from the shared data file, so search, comparison, and the calculator read the same values.
Qwen3.6-27B
Qwen/Qwen3.6-27B
DeepSeek V4-Pro
deepseek-v4-pro
DeepSeek-V4.1-Flash
deepseek-flash
Llama 4 Maverick
meta-llama/Llama-4-Maverick-17B-128E-Instruct
Llama 4 Scout
meta-llama/Llama-4-Scout-17B-16E
Phi-4
microsoft/phi-4
Mistral Large 3
mistral-large-2512
Mistral Medium 3.5
mistral-medium-3-5
Kimi K2.6
kimi-k2.6
Kimi K3
kimi-k3
GLM-5.2
glm-5.2
Qwen3.8-Flash
qwen3.8-flash
License, context, and published API fields remain attached to their sources. Hardware fit and deployment priority are editorial planning aids. Read the methodology →
Continue the decision
Do not stop at a list of names. Carry a candidate into comparison, cost planning, or migration checks.
Compare candidates
Put two or more models side by side using the same data fields.
02Price the workload
Turn input, output, and cache volumes into a monthly estimate.
03Copy the exact model ID
Every hosted identifier in one searchable list, with the provider page it was read from.
04Watch for changes
Review releases, retirements, and replacement routes before production.