Models
The converter keys on architectures[0] from a checkpoint’s
config.json. 52 models convert today, and any MoE
repository works as <user>/<model>.
-
DeepSeek-V2
DeepseekV2ForCausalLM -
DeepSeek-V2-Lite
DeepseekV2ForCausalLM -
DeepSeek-V3
DeepseekV3ForCausalLM -
DeepSeek-R1
DeepseekV3ForCausalLM -
Kimi-K2
DeepseekV3ForCausalLM -
Qwen3-MoE 30B-A3B
Qwen3MoeForCausalLM -
Qwen3-MoE 235B-A22B
Qwen3MoeForCausalLM -
Qwen2-MoE
Qwen2MoeForCausalLM -
Qwen3-Next
Qwen3NextForCausalLM -
Qwen3.5
Qwen3_5MoeForCausalLM -
Qwen3.8
Qwen3_5MoeForCausalLM -
Qwen 3.6
Qwen3_5MoeForConditionalGeneration -
Qwen3.5 multimodal
Qwen3_5MoeForConditionalGeneration -
Mixtral 8x7B
MixtralForCausalLM -
Mixtral 8x22B
MixtralForCausalLM -
OLMoE
OlmoeForCausalLM -
dots.llm1
Dots1ForCausalLM -
GLM-4.5
Glm4MoeForCausalLM -
GLM-4.6
Glm4MoeForCausalLM -
GLM-5.2
GlmMoeDsaForCausalLM -
gpt-oss-20B
GptOssForCausalLM -
gpt-oss-120B
GptOssForCausalLM -
Phi-3.5-MoE
PhiMoEForCausalLM -
Granite-MoE
GraniteMoeForCausalLM -
DBRX
DbrxForCausalLM -
Hunyuan-MoE
HunYuanMoEV1ForCausalLM -
Hunyuan V3
HYV3ForCausalLM -
Llama-4 Scout
Llama4ForConditionalGeneration -
Llama-4 Maverick
Llama4ForConditionalGeneration -
MiniMax-Text-01
MiniMaxText01ForCausalLM -
MiniMax-M1
MiniMaxM1ForCausalLM -
MiniMax-M2
MiniMaxM2ForCausalLM -
MiniMax-M3
MiniMaxM3SparseForCausalLM -
MiniMax-M3 multimodal
MiniMaxM3SparseForConditionalGeneration -
Grok 1
Grok1ForCausalLM, Grok1ModelForCausalLM, GrokForCausalLM -
Step 3.5 Flash
Step3p5ForCausalLM -
Step 3.7 Flash
Step3p5ForCausalLM -
Gemma 4
Gemma4ForCausalLM -
Gemma 4 multimodal
Gemma4ForConditionalGeneration -
Ling 1.0
BailingMoeForCausalLM -
Ling 2.0
BailingMoeV2ForCausalLM -
Ring 2.0
BailingMoeV2ForCausalLM -
Ling 2.5
BailingMoeV2_5ForCausalLM -
Ring 2.5
BailingMoeV2_5ForCausalLM -
Sarvam's MoE
SarvamMoEForCausalLM -
Snowflake Arctic
ArcticForCausalLM -
ERNIE 4.5 MoE
Ernie4_5_MoeForCausalLM -
Nemotron-H
NemotronHForCausalLM -
Nemotron 3.5 Lightning
NemotronHForCausalLM -
Kimi Linear
KimiLinearForCausalLM -
Kimi K3
KimiK3ForConditionalGeneration -
Mistral Large 3
MistralLarge3ForCausalLM
No architecture matches that.