Model techniques map

The models · 28 open-weight · 13 labs

28 models, ranked by how much they tell you.

The open-weight models the map covers. Entry is by the OpenRouter weekly token ranking plus explicit additions; a model leaving the ranking keeps its page, because the techniques its documentation described do not expire.

“Methods documented” measures how openly a lab writes about its model — a 200-method report and a 17-method model card can describe equally capable models.

ModelCreatorMethods documentedSourcesReleasedUsage rank
NVIDIA-Nemotron-3-Ultra-550B-A55BNVIDIA20872026-06-04#5
DeepSeek-V4.1-FlashDeepSeek20142026-09-10#2
Kimi K3Moonshot AI183102026-07-16—
Laguna-S-2.1Poolside16382026-07-21—
MiMo-V2.6-ProXiaomi14352026-09-22—
MiMo-V2.6-FlashXiaomi13652026-09-22#6
Qwen3.8-Flash-NextAlibaba (Qwen)12672026-08-26—
DeepSeek-V4-ProDeepSeek9852026-04-24—
Qwen3.5-397B-A17BAlibaba (Qwen)9882026-02-16—
gpt-oss-120bOpenAI8662025-08-05—
MiniMax-M3MiniMax8372026-06-01#12
DeepSeek-V4-FlashDeepSeek8242026-04-24#7
Gemma 4 31BGoogle DeepMind78102026-04-02—
MiMo-V2.5Xiaomi7782026-04-22#8
InklingThinking Machines Lab73112026-07-15—
NVIDIA-Nemotron-3.5-Lightning-30B-A3BNVIDIA7372026-08-11—
GLM-5.2Z.ai (Zhipu AI)7252026-06-16#11
DeepSeek-V3.2DeepSeek6842025-12-01—
GLM-5.3Z.ai (Zhipu AI)4662026-08-16#9
Qwen3.6-35B-A3BAlibaba (Qwen)4662026-04-15—
Step-3.7-FlashStepFun4262026-05-29—
MiMo-V2.5-ProXiaomi4162026-04-22—
Hy3Tencent (Hunyuan)3872026-07-06#10
Hy4-previewTencent (Hunyuan)3772026-08-28#3
GLM-5.3-FlashZ.ai (Zhipu AI)3352026-08-26#1
DeepSeek-V4-Flash-Vision-ExpDeepSeek2532026-08-21—
DeepSeek-V4-Flash-0731DeepSeek1742026-07-31#4
DeepSeek-V4-Pro-0813DeepSeek1332026-08-13—

Rank is the model's position among open-weight models in OpenRouter's token ranking for the week of 2026-09-21 (a partial week), the same week for every model; — means outside that week's top 12.