INTELLECT-3
Prime Intellect · GLM-4.5 · open weights
- GQA
- RoPE
- Partial RoPE
- Pre-norm
- MoE
- Shared expert
- Dense first layers
- MTP
Facts and where they come from
| Released | 2025-11 | config.jsonconfig.jsonHugging Face repository creation date (api.createdAt) |
|---|---|---|
| Licence | mit | config.jsonconfig.jsonREADME metadata: license |
| Total parameters | not disclosed | not disclosed |
| Active parameters | not disclosed | not disclosed |
| Context length | 128K tokens | config.jsonconfig.jsonmax_position_embeddings |
| Norm placement | pre | codemodelling codetransformers 5.18.0 / repo modelling code for glm4_moe: input_layernorm before attention, post_attention_layernorm before the MLP |
| Norm type | RMSNorm | codemodelling codetransformers 5.18.0 / repo modelling code for glm4_moe: input_layernorm before attention, post_attention_layernorm before the MLP |
| QK-norm | no | config.jsonconfig.jsonuse_qk_norm |
| Positional encoding | RoPE on 50% of each head | config.jsonconfig.jsonpartial_rotary_factor |
| Parallel attention and MLP | no | codemodelling codetransformers 5.18.0 / repo modelling code for glm4_moe: input_layernorm before attention, post_attention_layernorm before the MLP |
Architecture, drawn from the data
GQA 96q/8kv. Each column is one layer: its token mixer above, its feed-forward block below. Paler columns reuse another layer’s keys and values.
Modelled costs
From the cost model, batch size 1. Totals the lab states are in the table above; differences come from rounding, from what a lab counts, or from parts the model does not describe (listed on the about page).
| Parameters (modelled) | 107B |
|---|---|
| Active per token (modelled) | 13.4B |
| Without embeddings and output head | 106B total, 12.2B active |
| Multi-token-prediction layers (extra) | 2.38B |
| Published weights (Hugging Face count) | 107B |
| KV cache per token, BF16 (layers that grow with context) | 184 KiB |
| KV cache + state at 128K tokens, BF16 | 23 GiB |
| Decode FLOPs per token at 4K context | 34.9 GFLOP |
| Prefill FLOPs for a 4K prompt | 119 TFLOP |
KV cache against context
Every architecture field
| Field | Value | Source |
|---|---|---|
| d_model | 4,096 | config.jsonconfig.jsonhidden_size |
| vocab | 151,552 | config.jsonconfig.jsonvocab_size |
| tied_embeddings | false | config.jsonconfig.jsontie_word_embeddings |
| mixers.full.type | attn | codemodelling codeattention block |
| mixers.full.heads | 96 | config.jsonconfig.jsonnum_attention_heads |
| mixers.full.kv_heads | 8 | config.jsonconfig.jsonnum_key_value_heads |
| mixers.full.head_dim | 128 | config.jsonconfig.jsonhead_dim |
| mixers.full.bias | true | codemodelling codeattention_bias: true |
| ffns.dense.type | dense | codemodelling codeMLP block |
| ffns.dense.d_ff | 10,944 | config.jsonconfig.jsonintermediate_size |
| ffns.dense.gated | true | codemodelling codeMLP: gated (SwiGLU/GeGLU) |
| ffns.moe.type | moe | codemodelling codeMoE block |
| ffns.moe.experts | 128 | config.jsonconfig.jsonn_routed_experts |
| ffns.moe.active | 8 | config.jsonconfig.jsonnum_experts_per_tok |
| ffns.moe.d_expert | 1,408 | config.jsonconfig.jsonmoe_intermediate_size |
| ffns.moe.gated | true | codemodelling codeexperts are gated MLPs |
| ffns.moe.shared | 1 | config.jsonconfig.jsonn_shared_experts |
| ffns.moe.d_shared | 1,408 | config.jsonconfig.jsonmoe_intermediate_size |
| layout | 1× full/dense · 45× full/moe | config.jsonconfig.jsonnum_hidden_layers, first_k_dense_replace |
| mtp_layers | 1 | config.jsonconfig.jsonnum_nextn_predict_layers |
Sources
- config.json @ ff39d4a
- model card
- announcement
- modelling code · transformers 5.18.0 modelling code, or the model repository's own modelling file at the pinned revision
Listed in the LLM Architecture Gallery checklist as “INTELLECT-3 (106B)” (name only; see about).