Motif-3 Beta debuts as a 314B MoE with 13B active parameters and 256K context
eliebakouch · x · 2026-07-21
Motif Technologies has released a preview/beta checkpoint of Motif-3, a large MoE language model with 314B total parameters and 13B active per token.
- The model card says the repo is an intermediate checkpoint; the final checkpoint is coming soon.
- Key specs highlighted in the screenshot:
- 256K context length (262,144 tokens)
- 384 experts, with 8 activated per token plus 1 shared expert
- Multilingual, general-purpose, with text generation, transformers, and safetensors support
- The tweet adds that Motif’s internal research includes a per-expert activation function (“polynorm”) and a variant of differential attention (“GDLA”), plus modified mHC.
- The post claims the model performs on par with larger systems such as MiniMax M3 and DeepSeek V4 Pro.
Related event: Korea's Motif Releases Motif-3 Beta with 314B Parameters on Hugging Face(4 posts)→
More from Models
- Anthropic's Terra returns with better writing and 2X usage vs 6.1 Sol — weswinder · 2026-10-27
- User catches model failing to parse 400-paper file, citing stale news sources instead — jungofthewon · 2026-10-27
- Odd eval result: Gemini 4.1 Flash gets more accurate when asked to control its CoT — dfrsrchtwts · 2026-10-27
- Poster surprised that v4.1 flash gets more accurate when asked to control its CoT — dfrsrchtwts · 2026-10-27
- Dev claims Anthropic is baking 'sabotage' into model behavior for 'wrong' engineering — cephaloform · 2026-10-27
- Gemini fails at listing unread emails, tells user to search Gmail manually — gethackteam · 2026-10-27