Supported models
Sonar reads the architectures field from a model’s config.json file. The value must match an architecture in these tables. A compatible Transformers implementation can also use the Transformers backend.
The current registry contains 369 architecture entries in the user-facing categories below.
Text generation
| Architecture | Implementation |
|---|---|
AXK1ForCausalLM | AXK1:AXK1ForCausalLM |
AfmoeForCausalLM | afmoe:AfmoeForCausalLM |
ApertusForCausalLM | apertus:ApertusForCausalLM |
ArceeForCausalLM | arcee:ArceeForCausalLM |
ArcticForCausalLM | arctic:ArcticForCausalLM |
BailingMoeForCausalLM | bailing_moe:BailingMoeForCausalLM |
BailingMoeV2ForCausalLM | bailing_moe:BailingMoeV2ForCausalLM |
BailingMoeV2_5ForCausalLM | bailing_moe_linear:BailingMoeV25ForCausalLM |
BloomForCausalLM | bloom:BloomForCausalLM |
ChatGLMForConditionalGeneration | chatglm:ChatGLMForCausalLM |
ChatGLMModel | chatglm:ChatGLMForCausalLM |
Cohere2ForCausalLM | commandr:CohereForCausalLM |
Cohere2MoeForCausalLM | cohere2_moe:Cohere2MoeForCausalLM |
CohereForCausalLM | commandr:CohereForCausalLM |
CohereMoeForCausalLM | cohere_moe:CohereMoeForCausalLM |
CwmForCausalLM | llama:LlamaForCausalLM |
DbrxForCausalLM | dbrx:DbrxForCausalLM |
DeciLMForCausalLM | nemotron_nas:DeciLMForCausalLM |
DeepseekForCausalLM | deepseek_v2:DeepseekForCausalLM |
DeepseekV2ForCausalLM | deepseek_v2:DeepseekV2ForCausalLM |
DeepseekV32ForCausalLM | deepseek_v2:DeepseekV3ForCausalLM |
DeepseekV3ForCausalLM | deepseek_v2:DeepseekV3ForCausalLM |
DeepseekV4ForCausalLM | aphrodite.models.deepseek_v4:DeepseekV4ForCausalLM |
Ernie4_5ForCausalLM | ernie45:Ernie4_5ForCausalLM |
Ernie4_5_MoeForCausalLM | ernie45_moe:Ernie4_5_MoeForCausalLM |
Exaone4ForCausalLM | exaone4:Exaone4ForCausalLM |
ExaoneForCausalLM | exaone:ExaoneForCausalLM |
ExaoneMoEForCausalLM | exaone_moe:ExaoneMoeForCausalLM |
Fairseq2LlamaForCausalLM | fairseq2_llama:Fairseq2LlamaForCausalLM |
FalconForCausalLM | falcon:FalconForCausalLM |
FalconH1ForCausalLM | falcon_h1:FalconH1ForCausalLM |
FalconMambaForCausalLM | mamba:MambaForCausalLM |
FlexOlmoForCausalLM | flex_olmo:FlexOlmoForCausalLM |
GPT2LMHeadModel | gpt2:GPT2LMHeadModel |
GPTJForCausalLM | gpt_j:GPTJForCausalLM |
GPTNeoXForCausalLM | gpt_neox:GPTNeoXForCausalLM |
Gemma2ForCausalLM | gemma2:Gemma2ForCausalLM |
Gemma3ForCausalLM | gemma3:Gemma3ForCausalLM |
Gemma3nForCausalLM | gemma3n:Gemma3nForCausalLM |
Gemma4ForCausalLM | gemma4:Gemma4ForCausalLM |
GemmaForCausalLM | gemma:GemmaForCausalLM |
Glm4ForCausalLM | glm4:Glm4ForCausalLM |
Glm4MoeForCausalLM | glm4_moe:Glm4MoeForCausalLM |
Glm4MoeLiteForCausalLM | glm4_moe_lite:Glm4MoeLiteForCausalLM |
GlmForCausalLM | glm:GlmForCausalLM |
GlmMoeDsaForCausalLM | deepseek_v2:GlmMoeDsaForCausalLM |
GptOssForCausalLM | gpt_oss:GptOssForCausalLM |
GraniteForCausalLM | granite:GraniteForCausalLM |
GraniteMoeForCausalLM | granitemoe:GraniteMoeForCausalLM |
GraniteMoeHybridForCausalLM | granitemoehybrid:GraniteMoeHybridForCausalLM |
GraniteMoeSharedForCausalLM | granitemoeshared:GraniteMoeSharedForCausalLM |
GritLM | gritlm:GritLM |
HCXVisionForCausalLM | hyperclovax_vision:HCXVisionForCausalLM |
HCXVisionV2ForCausalLM | hyperclovax_vision_v2:HCXVisionV2ForCausalLM |
HYV3ForCausalLM | hy_v3:HYV3ForCausalLM |
HrmTextForCausalLM | hrm_text:HrmTextForCausalLM |
HunYuanDenseV1ForCausalLM | hunyuan_v1:HunYuanDenseV1ForCausalLM |
HunYuanMoEV1ForCausalLM | hunyuan_v1:HunYuanMoEV1ForCausalLM |
HyperCLOVAXForCausalLM | hyperclovax:HyperCLOVAXForCausalLM |
IQuestCoderForCausalLM | llama:LlamaForCausalLM |
IQuestLoopCoderForCausalLM | iquest_loopcoder:IQuestLoopCoderForCausalLM |
InklingForCausalLM | aphrodite.models.inkling:InklingForCausalLM |
InklingForConditionalGeneration | aphrodite.models.inkling:InklingForConditionalGeneration |
InternLM2ForCausalLM | internlm2:InternLM2ForCausalLM |
InternLM3ForCausalLM | llama:LlamaForCausalLM |
Jais2ForCausalLM | jais2:Jais2ForCausalLM |
JambaForCausalLM | jamba:JambaForCausalLM |
KimiLinearForCausalLM | kimi_linear:KimiLinearForCausalLM |
LLaMAForCausalLM | llama:LlamaForCausalLM |
LagunaForCausalLM | laguna:LagunaForCausalLM |
Lfm2ForCausalLM | lfm2:Lfm2ForCausalLM |
Lfm2MoeForCausalLM | lfm2_moe:Lfm2MoeForCausalLM |
Llama4ForCausalLM | llama4:Llama4ForCausalLM |
LlamaForCausalLM | llama:LlamaForCausalLM |
LongcatFlashForCausalLM | longcat_flash:LongcatFlashForCausalLM |
LongcatFlashNgramForCausalLM | longcat_flash_ngram:LongcatFlashNgramForCausalLM |
MPTForCausalLM | mpt:MPTForCausalLM |
Mamba2ForCausalLM | mamba2:Mamba2ForCausalLM |
MambaForCausalLM | mamba:MambaForCausalLM |
MellumForCausalLM | mellum:MellumForCausalLM |
MiMoForCausalLM | mimo:MiMoForCausalLM |
MiMoV2FlashForCausalLM | mimo_v2:MiMoV2FlashForCausalLM |
MiMoV2ForCausalLM | mimo_v2:MiMoV2ForCausalLM |
MiniCPM3ForCausalLM | minicpm3:MiniCPM3ForCausalLM |
MiniCPMForCausalLM | minicpm:MiniCPMForCausalLM |
MiniMaxM2ForCausalLM | minimax_m2:MiniMaxM2ForCausalLM |
MiniMaxM3SparseForCausalLM | aphrodite.models.minimax_m3:MiniMaxM3SparseForCausalLM |
Ministral3ForCausalLM | mistral:MistralForCausalLM |
MistralForCausalLM | mistral:MistralForCausalLM |
MistralLarge3ForCausalLM | mistral_large_3:MistralLarge3ForCausalLM |
MixtralForCausalLM | mixtral:MixtralForCausalLM |
MptForCausalLM | mpt:MPTForCausalLM |
NemotronForCausalLM | nemotron:NemotronForCausalLM |
NemotronHForCausalLM | nemotron_h:NemotronHForCausalLM |
NemotronHPuzzleForCausalLM | nemotron_h:NemotronHForCausalLM |
OPTForCausalLM | opt:OPTForCausalLM |
Olmo3ForCausalLM | olmo3:Olmo3ForCausalLM |
OlmoHybridForCausalLM | olmo_hybrid:OlmoHybridForCausalLM |
OlmoeForCausalLM | olmoe:OlmoeForCausalLM |
OrionForCausalLM | orion:OrionForCausalLM |
PanguEmbeddedForCausalLM | openpangu:PanguEmbeddedForCausalLM |
PanguProMoEV2ForCausalLM | openpangu:PanguProMoEV2ForCausalLM |
PanguUltraMoEForCausalLM | openpangu:PanguUltraMoEForCausalLM |
Param2MoEForCausalLM | param2moe:Param2MoEForCausalLM |
Phi3ForCausalLM | phi3:Phi3ForCausalLM |
PhiForCausalLM | phi:PhiForCausalLM |
PhiMoEForCausalLM | phimoe:PhiMoEForCausalLM |
Plamo3ForCausalLM | plamo3:Plamo3ForCausalLM |
Qwen2ForCausalLM | qwen2:Qwen2ForCausalLM |
Qwen2MoeForCausalLM | qwen2_moe:Qwen2MoeForCausalLM |
Qwen3ForCausalLM | qwen3:Qwen3ForCausalLM |
Qwen3MoeForCausalLM | qwen3_moe:Qwen3MoeForCausalLM |
Qwen3NextForCausalLM | qwen3_next:Qwen3NextForCausalLM |
RWForCausalLM | falcon:FalconForCausalLM |
Rnj1ForCausalLM | rnj1:Rnj1ForCausalLM |
SarvamMLAForCausalLM | sarvam:SarvamMLAForCausalLM |
SarvamMoEForCausalLM | sarvam:SarvamMoEForCausalLM |
SeedOssForCausalLM | seed_oss:SeedOssForCausalLM |
SolarForCausalLM | solar:SolarForCausalLM |
StableLMEpochForCausalLM | stablelm:StablelmForCausalLM |
StableLmForCausalLM | stablelm:StablelmForCausalLM |
Step1ForCausalLM | step1:Step1ForCausalLM |
Step3TextForCausalLM | step3_text:Step3TextForCausalLM |
Step3p5ForCausalLM | step3p5:Step3p5ForCausalLM |
TeleChat2ForCausalLM | telechat2:TeleChat2ForCausalLM |
TeleChat3ForCausalLM | llama:LlamaForCausalLM |
TeleFLMForCausalLM | teleflm:TeleFLMForCausalLM |
Zamba2ForCausalLM | zamba2:Zamba2ForCausalLM |
Embedding
| Architecture | Implementation |
|---|---|
BertForMaskedLM | bert:BertForMaskedLM |
BertModel | bert:BertEmbeddingModel |
BertSpladeSparseEmbeddingModel | bert:BertSpladeSparseEmbeddingModel |
BgeM3EmbeddingModel | roberta:BgeM3EmbeddingModel |
CLIPModel | clip:CLIPEmbeddingModel |
ColPaliForRetrieval | colpali:ColPaliModel |
CwmForCausalLM | llama:LlamaForCausalLM |
DeciLMForCausalLM | nemotron_nas:DeciLMForCausalLM |
Gemma2Model | gemma2:Gemma2ForCausalLM |
Gemma3TextModel | gemma3:Gemma3Model |
GlmForCausalLM | glm:GlmForCausalLM |
GritLM | gritlm:GritLM |
GteModel | bert_with_rope:SnowflakeGteNewModel |
GteNewModel | bert_with_rope:GteNewModel |
IQuestCoderForCausalLM | llama:LlamaForCausalLM |
InternLM3ForCausalLM | llama:LlamaForCausalLM |
JinaEmbeddingsV5Model | jina:JinaEmbeddingsV5Model |
LLaMAForCausalLM | llama:LlamaForCausalLM |
LlamaBidirectionalModel | llama:LlamaBidirectionalModel |
LlamaForCausalLM | llama:LlamaForCausalLM |
LlamaModel | llama:LlamaForCausalLM |
LlamaNemotronVLModel | nemotron_vl:LlamaNemotronVLForEmbedding |
LlavaNextForConditionalGeneration | llava_next:LlavaNextForConditionalGeneration |
MistralModel | llama:LlamaForCausalLM |
ModernBertModel | modernbert:ModernBertModel |
NomicBertModel | bert_with_rope:NomicBertModel |
Phi3ForCausalLM | phi3:Phi3ForCausalLM |
Phi3VForCausalLM | phi3v:Phi3VForCausalLM |
PrithviGeoSpatialMAE | terratorch:Terratorch |
Qwen2ForCausalLM | qwen2:Qwen2ForCausalLM |
Qwen2Model | qwen2:Qwen2ForCausalLM |
Qwen2VLForConditionalGeneration | qwen2_vl:Qwen2VLForConditionalGeneration |
RobertaForMaskedLM | roberta:RobertaEmbeddingModel |
RobertaModel | roberta:RobertaEmbeddingModel |
SiglipModel | siglip:SiglipEmbeddingModel |
TeleChat2ForCausalLM | telechat2:TeleChat2ForCausalLM |
TeleChat3ForCausalLM | llama:LlamaForCausalLM |
Terratorch | terratorch:Terratorch |
VoyageQwen3BidirectionalEmbedModel | voyage:VoyageQwen3BidirectionalEmbedModel |
XLMRobertaModel | roberta:RobertaEmbeddingModel |
Late interaction
| Architecture | Implementation |
|---|---|
ColBERTJinaRobertaModel | colbert:ColBERTJinaRobertaModel |
ColBERTLfm2Model | colbert:ColBERTLfm2Model |
ColBERTModernBertModel | colbert:ColBERTModernBertModel |
ColModernVBertForRetrieval | colmodernvbert:ColModernVBertForRetrieval |
ColPaliForRetrieval | colpali:ColPaliModel |
ColQwen3 | colqwen3:ColQwen3Model |
ColQwen3_5 | colqwen3_5:ColQwen3_5Model |
HF_ColBERT | colbert:ColBERTModel |
JinaForRanking | jina:JinaForRanking |
OpsColQwen3Model | colqwen3:ColQwen3Model |
Qwen3VLNemotronEmbedModel | colqwen3:ColQwen3Model |
Reward
| Architecture | Implementation |
|---|---|
InternLM2ForRewardModel | internlm2:InternLM2ForRewardModel |
Qwen2ForProcessRewardModel | qwen2_rm:Qwen2ForProcessRewardModel |
Qwen2ForRewardModel | qwen2_rm:Qwen2ForRewardModel |
Token classification
| Architecture | Implementation |
|---|---|
BertForTokenClassification | bert:BertForTokenClassification |
ModernBertForTokenClassification | modernbert:ModernBertForTokenClassification |
OpenAIPrivacyFilterForTokenClassification | openai_privacy_filter:OpenAIPrivacyFilterForTokenClassification |
Qwen3ASRForcedAlignerForTokenClassification | qwen3_asr_forced_aligner:Qwen3ASRForcedAlignerForTokenClassification |
RobertaForTokenClassification | roberta:RobertaForTokenClassification |
XLMRobertaForTokenClassification | roberta:RobertaForTokenClassification |
Sequence classification
| Architecture | Implementation |
|---|---|
BertForSequenceClassification | bert:BertForSequenceClassification |
GPT2ForSequenceClassification | gpt2:GPT2ForSequenceClassification |
GteNewForSequenceClassification | bert_with_rope:GteNewForSequenceClassification |
JambaForSequenceClassification | jamba:JambaForSequenceClassification |
JinaVLForRanking | jina_vl:JinaVLForSequenceClassification |
LlamaBidirectionalForSequenceClassification | llama:LlamaBidirectionalForSequenceClassification |
LlamaNemotronVLForSequenceClassification | nemotron_vl:LlamaNemotronVLForSequenceClassification |
ModernBertForSequenceClassification | modernbert:ModernBertForSequenceClassification |
RobertaForSequenceClassification | roberta:RobertaForSequenceClassification |
XLMRobertaForSequenceClassification | roberta:RobertaForSequenceClassification |
Multimodal
| Architecture | Implementation |
|---|---|
AriaForConditionalGeneration | aria:AriaForConditionalGeneration |
AudioFlamingo3ForConditionalGeneration | audioflamingo3:AudioFlamingo3ForConditionalGeneration |
BagelForConditionalGeneration | bagel:BagelForConditionalGeneration |
BeeForConditionalGeneration | bee:BeeForConditionalGeneration |
Blip2ForConditionalGeneration | blip2:Blip2ForConditionalGeneration |
ChameleonForConditionalGeneration | chameleon:ChameleonForConditionalGeneration |
Cheers | cheers:CheersForConditionalGeneration |
CheersForConditionalGeneration | cheers:CheersForConditionalGeneration |
Cohere2VisionForConditionalGeneration | cohere2_vision:Cohere2VisionForConditionalGeneration |
CohereAsrForConditionalGeneration | cohere_asr:CohereAsrForConditionalGeneration |
Cosmos3EdgeForConditionalGeneration | cosmos3_edge:Cosmos3EdgeForConditionalGeneration |
Cosmos3ForConditionalGeneration | cosmos3:Cosmos3ForConditionalGeneration |
DeepseekOCR2ForCausalLM | deepseek_ocr2:DeepseekOCR2ForCausalLM |
DeepseekOCRForCausalLM | deepseek_ocr:DeepseekOCRForCausalLM |
DeepseekVLV2ForCausalLM | deepseek_vl2:DeepseekVLV2ForCausalLM |
DiffusionGemmaForBlockDiffusion | diffusion_gemma:DiffusionGemmaForConditionalGeneration |
DotsOCRForCausalLM | dots_ocr:DotsOCRForCausalLM |
Eagle2_5_VLForConditionalGeneration | eagle2_5_vl:Eagle2_5_VLForConditionalGeneration |
Ernie4_5_VLMoeForConditionalGeneration | ernie45_vl:Ernie4_5_VLMoeForConditionalGeneration |
Exaone4_5_ForConditionalGeneration | exaone4_5:Exaone4_5_ForConditionalGeneration |
FireRedASR2ForConditionalGeneration | fireredasr2:FireRedASR2ForConditionalGeneration |
FireRedLIDForConditionalGeneration | fireredlid:FireRedLIDForConditionalGeneration |
FunASRForConditionalGeneration | funasr:FunASRForConditionalGeneration |
FunAudioChatForConditionalGeneration | funaudiochat:FunAudioChatForConditionalGeneration |
GLM4VForCausalLM | glm4v:GLM4VForCausalLM |
Gemma3ForConditionalGeneration | gemma3_mm:Gemma3ForConditionalGeneration |
Gemma3nForConditionalGeneration | gemma3n_mm:Gemma3nForConditionalGeneration |
Gemma4ForConditionalGeneration | gemma4_mm:Gemma4ForConditionalGeneration |
Gemma4UnifiedForConditionalGeneration | gemma4_unified:Gemma4UnifiedForConditionalGeneration |
Glm4vForConditionalGeneration | glm4_1v:Glm4vForConditionalGeneration |
Glm4vMoeForConditionalGeneration | glm4_1v:Glm4vMoeForConditionalGeneration |
GlmAsrForConditionalGeneration | glmasr:GlmAsrForConditionalGeneration |
GlmOcrForConditionalGeneration | glm_ocr:GlmOcrForConditionalGeneration |
Granite4VisionForConditionalGeneration | granite4_vision:Granite4VisionForConditionalGeneration |
GraniteSpeechForConditionalGeneration | granite_speech:GraniteSpeechForConditionalGeneration |
GraniteSpeechPlusForConditionalGeneration | granite_speech_plus:GraniteSpeechPlusForConditionalGeneration |
H2OVLChatModel | h2ovl:H2OVLChatModel |
HfMoondream | moondream3:Moondream3ForCausalLM |
HunYuanVLForConditionalGeneration | hunyuan_vision:HunYuanVLForConditionalGeneration |
Idefics3ForConditionalGeneration | idefics3:Idefics3ForConditionalGeneration |
InternS1ForConditionalGeneration | interns1:InternS1ForConditionalGeneration |
InternS1ProForConditionalGeneration | interns1_pro:InternS1ProForConditionalGeneration |
InternS2PreviewForConditionalGeneration | interns2_preview:InternS2PreviewForConditionalGeneration |
InternVLChatModel | internvl:InternVLChatModel |
InternVLForConditionalGeneration | interns1:InternS1ForConditionalGeneration |
IsaacForConditionalGeneration | isaac:IsaacForConditionalGeneration |
KananaVForConditionalGeneration | kanana_v:KananaVForConditionalGeneration |
KeyeForConditionalGeneration | keye:KeyeForConditionalGeneration |
KeyeVL1_5ForConditionalGeneration | keye_vl1_5:KeyeVL1_5ForConditionalGeneration |
KimiK25ForConditionalGeneration | kimi_k25:KimiK25ForConditionalGeneration |
KimiVLForConditionalGeneration | kimi_vl:KimiVLForConditionalGeneration |
Lfm2VlForConditionalGeneration | lfm2_vl:Lfm2VLForConditionalGeneration |
LightOnOCRForConditionalGeneration | lightonocr:LightOnOCRForConditionalGeneration |
Llama4ForConditionalGeneration | mllama4:Llama4ForConditionalGeneration |
Llama_Nemotron_Nano_VL | nemotron_vl:LlamaNemotronVLChatModel |
LlavaForConditionalGeneration | llava:LlavaForConditionalGeneration |
LlavaNextForConditionalGeneration | llava_next:LlavaNextForConditionalGeneration |
LlavaNextVideoForConditionalGeneration | llava_next_video:LlavaNextVideoForConditionalGeneration |
LlavaOnevision2ForConditionalGeneration | llava_onevision2:LlavaOnevision2ForConditionalGeneration |
LlavaOnevisionForConditionalGeneration | llava_onevision:LlavaOnevisionForConditionalGeneration |
MiDashengLMModel | midashenglm:MiDashengLMModel |
MiMoV2OmniForCausalLM | mimo_v2_omni:MiMoV2OmniForCausalLM |
MiniCPMO | minicpmo:MiniCPMO |
MiniCPMV | minicpmv:MiniCPMV |
MiniCPMV4_6ForConditionalGeneration | minicpmv4_6:MiniCPMV4_6ForConditionalGeneration |
MiniMaxM3SparseForConditionalGeneration | aphrodite.models.minimax_m3:MiniMaxM3SparseForConditionalGeneration |
Mistral3ForConditionalGeneration | mistral3:Mistral3ForConditionalGeneration |
Molmo2ForConditionalGeneration | molmo2:Molmo2ForConditionalGeneration |
MolmoForCausalLM | molmo:MolmoForCausalLM |
Moondream3ForCausalLM | moondream3:Moondream3ForCausalLM |
MoonshotKimiaForCausalLM | kimi_audio:KimiAudioForConditionalGeneration |
MossAudioModel | moss_audio:MossAudioModel |
MossTranscribeDiarizeForConditionalGeneration | moss_transcribe_diarize:MossTranscribeDiarizeForConditionalGeneration |
NVLM_D | nvlm_d:NVLM_D_Model |
NemotronH_Nano_Omni_Reasoning_V3 | nano_nemotron_vl:NemotronH_Nano_VL_V2 |
NemotronH_Nano_VL_V2 | nano_nemotron_vl:NemotronH_Nano_VL_V2 |
NemotronH_Super_Omni_Reasoning_V3 | nano_nemotron_vl:NemotronH_Nano_VL_V2 |
NemotronParseForConditionalGeneration | nemotron_parse:NemotronParseForConditionalGeneration |
OpenCUAForConditionalGeneration | opencua:OpenCUAForConditionalGeneration |
OpenPanguVLForConditionalGeneration | openpangu_vl:OpenPanguVLForConditionalGeneration |
OpenVLAForActionPrediction | openvla:OpenVLAForActionPrediction |
Ovis | ovis:Ovis |
Ovis2_5 | ovis2_5:Ovis2_5 |
Ovis2_6ForCausalLM | ovis2_5:Ovis2_5 |
Ovis2_6_MoeForCausalLM | ovis2_5:Ovis2_5 |
PaddleOCRVLForConditionalGeneration | paddleocr_vl:PaddleOCRVLForConditionalGeneration |
PaliGemmaForConditionalGeneration | paligemma:PaliGemmaForConditionalGeneration |
Phi3VForCausalLM | phi3v:Phi3VForCausalLM |
Phi4ForCausalLMV | phi4siglip:Phi4ForCausalLMV |
Phi4MMForCausalLM | phi4mm:Phi4MMForCausalLM |
PixtralForConditionalGeneration | pixtral:PixtralForConditionalGeneration |
QianfanOCRForConditionalGeneration | qianfan_ocr:QianfanOCRForConditionalGeneration |
Qwen2AudioForConditionalGeneration | qwen2_audio:Qwen2AudioForConditionalGeneration |
Qwen2VLForConditionalGeneration | qwen2_vl:Qwen2VLForConditionalGeneration |
Qwen2_5OmniForConditionalGeneration | qwen2_5_omni_thinker:Qwen2_5OmniThinkerForConditionalGeneration |
Qwen2_5OmniModel | qwen2_5_omni_thinker:Qwen2_5OmniThinkerForConditionalGeneration |
Qwen2_5_VLForConditionalGeneration | qwen2_5_vl:Qwen2_5_VLForConditionalGeneration |
Qwen3ASRForConditionalGeneration | qwen3_asr:Qwen3ASRForConditionalGeneration |
Qwen3ASRRealtimeGeneration | qwen3_asr_realtime:Qwen3ASRRealtimeGeneration |
Qwen3OmniMoeForConditionalGeneration | qwen3_omni_moe_thinker:Qwen3OmniMoeThinkerForConditionalGeneration |
Qwen3VLForConditionalGeneration | qwen3_vl:Qwen3VLForConditionalGeneration |
Qwen3VLMoeForConditionalGeneration | qwen3_vl_moe:Qwen3VLMoeForConditionalGeneration |
Qwen3_5ForConditionalGeneration | qwen3_5:Qwen3_5ForConditionalGeneration |
Qwen3_5MoeForConditionalGeneration | qwen3_5:Qwen3_5MoeForConditionalGeneration |
RForConditionalGeneration | rvl:RForConditionalGeneration |
SkyworkR1VChatModel | skyworkr1v:SkyworkR1VChatModel |
SmolVLMForConditionalGeneration | smolvlm:SmolVLMForConditionalGeneration |
Step3VLForConditionalGeneration | step3_vl:Step3VLForConditionalGeneration |
Step3p7ForConditionalGeneration | step3p7:Step3p7ForConditionalGeneration |
StepVLForConditionalGeneration | step_vl:StepVLForConditionalGeneration |
UltravoxModel | ultravox:UltravoxModel |
UnlimitedOCRForCausalLM | unlimited_ocr:UnlimitedOCRForCausalLM |
VoxtralForConditionalGeneration | voxtral:VoxtralForConditionalGeneration |
VoxtralRealtimeGeneration | voxtral_realtime:VoxtralRealtimeGeneration |
WhisperForConditionalGeneration | whisper:WhisperForConditionalGeneration |
Speculative decoding
| Architecture | Implementation |
|---|---|
BailingMoeV25MTPModel | bailing_moe_mtp:BailingMoeV25MTPModel |
DFlashDraftModel | qwen3_dflash:DFlashQwen3ForCausalLM |
DFlashLagunaForCausalLM | laguna_dflash:DFlashLagunaForCausalLM |
DSparkDraftModel | aphrodite.models.deepseek_v4:DSparkDeepseekV4ForCausalLM |
DeepSeekMTPModel | deepseek_mtp:DeepSeekMTP |
DeepSeekV4MTPModel | aphrodite.models.deepseek_v4:DeepSeekV4MTP |
Eagle3DeepseekV2ForCausalLM | deepseek_eagle3:Eagle3DeepseekV2ForCausalLM |
Eagle3DeepseekV3ForCausalLM | deepseek_eagle3:Eagle3DeepseekV2ForCausalLM |
Eagle3LlamaForCausalLM | llama_eagle3:Eagle3LlamaForCausalLM |
Eagle3MiniMaxM2ForCausalLM | llama_eagle3:Eagle3LlamaForCausalLM |
Eagle3Qwen2_5vlForCausalLM | llama_eagle3:Eagle3LlamaForCausalLM |
Eagle3Qwen3ForCausalLM | qwen3_eagle3:Eagle3Qwen3ForCausalLM |
Eagle3Qwen3vlForCausalLM | llama_eagle3:Eagle3LlamaForCausalLM |
EagleCohereForCausalLM | cohere_eagle:EagleCohereForCausalLM |
EagleDeepSeekMTPModel | deepseek_eagle:EagleDeepseekV3ForCausalLM |
EagleLlama4ForCausalLM | llama4_eagle:EagleLlama4ForCausalLM |
EagleLlamaForCausalLM | llama_eagle:EagleLlamaForCausalLM |
EagleMiniCPMForCausalLM | minicpm_eagle:EagleMiniCPMForCausalLM |
EagleMistralForCausalLM | mistral_eagle:EagleMistralForCausalLM |
EagleMistralLarge3ForCausalLM | mistral_large_3_eagle:EagleMistralLarge3ForCausalLM |
ErnieMTPModel | ernie_mtp:ErnieMTP |
Exaone4_5_MTP | exaone4_5_mtp:Exaone4_5_MTP |
ExaoneMoeMTP | exaone_moe_mtp:ExaoneMoeMTP |
ExtractHiddenStatesModel | extract_hidden_states:ExtractHiddenStatesModel |
Gemma4DSparkModel | gemma4_dspark:Gemma4DSparkForCausalLM |
Gemma4MTPModel | gemma4_mtp:Gemma4MTP |
Glm4MoeLiteMTPModel | glm4_moe_lite_mtp:Glm4MoeLiteMTP |
Glm4MoeMTPModel | glm4_moe_mtp:Glm4MoeMTP |
GlmOcrMTPModel | glm_ocr_mtp:GlmOcrMTP |
HYV3MTPModel | hy_v3_mtp:HYV3MTP |
InklingMTPModel | aphrodite.models.inkling:InklingMTP |
LlamaForCausalLMEagle3 | llama_eagle3:Eagle3LlamaForCausalLM |
LongCatFlashMTPModel | longcat_flash_mtp:LongCatFlashMTP |
MedusaModel | medusa:Medusa |
MiMoMTPModel | mimo_mtp:MiMoMTP |
MiMoV2MTPModel | mimo_v2_mtp:MiMoV2MTP |
MiMoV2OmniMTPModel | mimo_v2_mtp:MiMoV2OmniMTP |
MiniMaxM3MTP | aphrodite.models.minimax_m3:MiniMaxM3MTP |
NemotronHMTPModel | nemotron_h_mtp:NemotronHMTP |
OpenPanguMTPModel | openpangu_mtp:OpenPanguMTP |
PEagleDraftModel | llama_eagle3:Eagle3LlamaForCausalLM |
PeagleLlamaForCausalLM | llama_eagle3:Eagle3LlamaForCausalLM |
PeagleQwen3ForCausalLM | qwen3_eagle3:Eagle3Qwen3ForCausalLM |
Qwen3DSparkModel | qwen3_dspark:Qwen3DSparkForCausalLM |
Qwen3NextMTP | qwen3_next_mtp:Qwen3NextMTP |
Qwen3_5MTP | qwen3_5_mtp:Qwen3_5MTP |
Qwen3_5MoeMTP | qwen3_5_mtp:Qwen3_5MoeMTP |
Step3p5MTP | step3p5_mtp:Step3p5MTP |
Transformers backend
| Architecture | Implementation |
|---|---|
Emu3ForConditionalGeneration | transformers:TransformersMultiModalForCausalLM |
GPTBigCodeForCausalLM | transformers:TransformersForCausalLM |
Olmo2ForCausalLM | transformers:TransformersForCausalLM |
OlmoForCausalLM | transformers:TransformersForCausalLM |
SmolLM3ForCausalLM | transformers:TransformersForCausalLM |
Starcoder2ForCausalLM | transformers:TransformersForCausalLM |
VaultGemmaForCausalLM | transformers:TransformersForCausalLM |
VibeVoiceAsrForConditionalGeneration | transformers:TransformersMultiModalForCausalLM |