Qwen · 2025-04 · Dense
Qwen3 32B
Qwen3 32B is a grouped-query attention (GQA) transformer released by Qwen in 2025-04, with 64 layers, hidden size 5120 and a context window of 40,960 tokens.
Layer stack
Attention ×64 · Layers 64
Key facts
| Family | Qwen |
|---|---|
| Released | 2025-04 |
| Params | 32.1 B |
| Context | 40,960 tokens |
| Attention | GQA (64:8) |
| Layers | 64 |
| Hidden | 5,120 |
| Heads | 64 |
| Vocab | 151,936 |
| Position | RoPE |
| Norm | RMSNorm |
| Activation | silu |
| Dtype | bf16 |
| Architecture class | Qwen3ForCausalLM |
Architecture overview
Attention
grouped-query attention (GQA) — 64 q-heads / 8 kv-heads. RoPE θ=1,000,000.
Feed-forward / MoE
It is a dense model: all 32.1B parameters participate in computing every token.
Field-level comparison
Compared against the previous model of the same lab; where none exists, against the structurally closest model. The ratio column is this model divided by the comparison model.
Full-field comparison vs predecessor Qwen3 30B A3B
| Model | Qwen3 32B | Qwen3 30B A3B | Ratio |
|---|---|---|---|
| model_type | qwen3 | qwen3_moe | ≠ |
| architectures | Qwen3ForCausalLM | Qwen3MoeForCausalLM | ≠ |
| hidden_size | 5120 | 2048 | ×2.50 |
| num_hidden_layers | 64 | 48 | ×1.33 |
| num_attention_heads | 64 | 32 | ×2.00 |
| num_key_value_heads | 8 | 4 | ×2.00 |
| head_dim | 128 | 128 | ≈1 |
| intermediate_size | 25600 | 6144 | ×4.17 |
| hidden_act | silu | silu | ＝ |
| num_experts_per_tok | — | 8 | |
| moe_intermediate_size | — | 768 | |
| rope_theta | 1000000 | 10000000 | ×0.100 |
| max_position_embeddings | 40960 | 262144 | ×0.156 |
| rms_norm_eps | 0.000001 | 0.000001 | ≈1 |
| vocab_size | 151936 | 151936 | ≈1 |
| torch_dtype | bfloat16 | bfloat16 | ＝ |
| attention_bias | false | false | ＝ |
| attention_dropout | 0 | 0 | ＝ |
| bos_token_id | 151643 | 151643 | ≈1 |
| eos_token_id | 151645 | 151645 | ≈1 |
| initializer_range | 0.02 | 0.02 | ≈1 |
| max_window_layers | 64 | 48 | ×1.33 |
| tie_word_embeddings | false | false | ＝ |
| transformers_version | 4.51.0 | 4.51.0 | ＝ |
| use_cache | true | true | ＝ |
| use_sliding_window | false | false | ＝ |
Most similar architectures
- Qwen3.6 27B (Qwen, 2026-04) 0.998
- Qwen3.8 27B (Qwen, 2026-08) 0.998
- Llama 3.1 70B (Meta, 2024-07) 0.995
- Qwen2 72B (Qwen, 2024-06) 0.995
- Qwen2.5 72B (Qwen, 2024-09) 0.995
- Qwen3 8B (Qwen, 2025-04) 0.992
Similarity values range from 0 (no shared categorical features) to 1 (identical profiles).
Raw config fields
24 fields
| architectures | Qwen3ForCausalLM |
|---|---|
| attention_bias | false |
| attention_dropout | 0 |
| bos_token_id | 151643 |
| eos_token_id | 151645 |
| head_dim | 128 |
| hidden_act | silu |
| hidden_size | 5120 |
| initializer_range | 0.02 |
| intermediate_size | 25600 |
| max_position_embeddings | 40960 |
| max_window_layers | 64 |
| model_type | qwen3 |
| num_attention_heads | 64 |
| num_hidden_layers | 64 |
| num_key_value_heads | 8 |
| rms_norm_eps | 0.000001 |
| rope_theta | 1000000 |
| tie_word_embeddings | false |
| torch_dtype | bfloat16 |
| transformers_version | 4.51.0 |
| use_cache | true |
| use_sliding_window | false |
| vocab_size | 151936 |