N Neurarch Try it See it Changes Bot Pricing API Open app

Comparisons / LLaMA-3 Block vs DeepSeek-V3

LLaMA-3 Block vs DeepSeek-V3

What a frontier open model adds to the block everyone started from.

DeepSeek-V3 has 666B more parameters than LLaMA-3 Block: 364 layers added, 2 removed, 5 changed.

Baseline

LLaMA-3 Block

Layers
8
Parameters
703M
Input
1 × 2048
Output
1 × 2048 × 4096
Forward-passes
yes
Est. train cost
$14.27
Compared

DeepSeek-V3

Layers
370
Parameters
666B
Input
1 × 163840
Output
1 × 163840 × 129280
Forward-passes
yes
Est. train cost
$670,780

The deltas

Every number is DeepSeek-V3 relative to LLaMA-3 Block.

Parameters
+666B (948× the size)
Layers
+362
Added
364
Removed
2
Changed
5
Unchanged
3

Which GPUs each one fits

Each side is measured at its own declared input (1 × 2048 against 1 × 163840). Both columns are right about their own model; the difference between them is not a fact about the designs.

GPULLaMA-3 BlockDeepSeek-V3
T4 16GBnono
A100 40GBfitsno
H100 80GBfitsno

Layer by layer

Aligned in topological order. 3 of 374 rows are the same layer with the same parameters.

Show all 374 rows
LLaMA-3 BlockDeepSeek-V3
LayerParamsOutputLayerParamsOutput
1changed
shape
tokens
Input
1 × 2048Input
Input
1 × 163840
2removedrope
Rope
—
3changed
numEmbeddings, embeddingDim, vocabSize, maxSeqLen
embed
Embedding
525M1 × 2048 × 4096Embedding
Embedding
927M1 × 163840 × 7168
4added—RoPE (decoupled)
Rope
1 × 163840 × 7168
5changed
normalizedShape
attn_norm
Rms Norm
4.1K1 × 2048 × 4096RMSNorm_1_1
Rms Norm
7.2K1 × 163840 × 7168
6removedgqa
Grouped Query Attention
42M1 × 2048 × 4096—
7added—MLA_1
Mla
110M1 × 163840 × 7168
8added—Add_1_attn
Add
1 × 163840 × 7168
9added—RMSNorm_1_2
Rms Norm
7.2K1 × 163840 × 7168
10added—FFN_1
Swiglu
396M1 × 163840 × 7168
11added—Add_1_ffn
Add
1 × 163840 × 7168
12added—RMSNorm_2_1
Rms Norm
7.2K1 × 163840 × 7168
13added—MLA_2
Mla
110M1 × 163840 × 7168
14added—Add_2_attn
Add
1 × 163840 × 7168
15added—RMSNorm_2_2
Rms Norm
7.2K1 × 163840 × 7168
16added—FFN_2
Swiglu
396M1 × 163840 × 7168
17added—Add_2_ffn
Add
1 × 163840 × 7168
18added—RMSNorm_3_1
Rms Norm
7.2K1 × 163840 × 7168
19added—MLA_3
Mla
110M1 × 163840 × 7168
20sameresidual_1
Add
1 × 2048 × 4096Add_3_attn
Add
1 × 163840 × 7168
21changed
normalizedShape
ffn_norm
Rms Norm
4.1K1 × 2048 × 4096RMSNorm_3_2
Rms Norm
7.2K1 × 163840 × 7168
22changed
embedDim, intermediateSize, dim
swiglu_ffn
Swiglu
135M1 × 2048 × 4096FFN_3
Swiglu
396M1 × 163840 × 7168
23added—Add_3_ffn
Add
1 × 163840 × 7168
24added—RMSNorm_4_1
Rms Norm
7.2K1 × 163840 × 7168
25added—MLA_4
Mla
110M1 × 163840 × 7168
26added—Add_4_attn
Add
1 × 163840 × 7168
27added—RMSNorm_4_2
Rms Norm
7.2K1 × 163840 × 7168
28added—MoE_4
Shared Expert Mo E
11B1 × 163840 × 7168
29added—Add_4_ffn
Add
1 × 163840 × 7168
30added—RMSNorm_5_1
Rms Norm
7.2K1 × 163840 × 7168
31added—MLA_5
Mla
110M1 × 163840 × 7168
32added—Add_5_attn
Add
1 × 163840 × 7168
33added—RMSNorm_5_2
Rms Norm
7.2K1 × 163840 × 7168
34added—MoE_5
Shared Expert Mo E
11B1 × 163840 × 7168
35added—Add_5_ffn
Add
1 × 163840 × 7168
36added—RMSNorm_6_1
Rms Norm
7.2K1 × 163840 × 7168
37added—MLA_6
Mla
110M1 × 163840 × 7168
38added—Add_6_attn
Add
1 × 163840 × 7168
39added—RMSNorm_6_2
Rms Norm
7.2K1 × 163840 × 7168
40added—MoE_6
Shared Expert Mo E
11B1 × 163840 × 7168
41added—Add_6_ffn
Add
1 × 163840 × 7168
42added—RMSNorm_7_1
Rms Norm
7.2K1 × 163840 × 7168
43added—MLA_7
Mla
110M1 × 163840 × 7168
44added—Add_7_attn
Add
1 × 163840 × 7168
45added—RMSNorm_7_2
Rms Norm
7.2K1 × 163840 × 7168
46added—MoE_7
Shared Expert Mo E
11B1 × 163840 × 7168
47added—Add_7_ffn
Add
1 × 163840 × 7168
48added—RMSNorm_8_1
Rms Norm
7.2K1 × 163840 × 7168
49added—MLA_8
Mla
110M1 × 163840 × 7168
50added—Add_8_attn
Add
1 × 163840 × 7168
51added—RMSNorm_8_2
Rms Norm
7.2K1 × 163840 × 7168
52added—MoE_8
Shared Expert Mo E
11B1 × 163840 × 7168
53added—Add_8_ffn
Add
1 × 163840 × 7168
54added—RMSNorm_9_1
Rms Norm
7.2K1 × 163840 × 7168
55added—MLA_9
Mla
110M1 × 163840 × 7168
56added—Add_9_attn
Add
1 × 163840 × 7168
57added—RMSNorm_9_2
Rms Norm
7.2K1 × 163840 × 7168
58added—MoE_9
Shared Expert Mo E
11B1 × 163840 × 7168
59added—Add_9_ffn
Add
1 × 163840 × 7168
60added—RMSNorm_10_1
Rms Norm
7.2K1 × 163840 × 7168
61added—MLA_10
Mla
110M1 × 163840 × 7168
62added—Add_10_attn
Add
1 × 163840 × 7168
63added—RMSNorm_10_2
Rms Norm
7.2K1 × 163840 × 7168
64added—MoE_10
Shared Expert Mo E
11B1 × 163840 × 7168
65added—Add_10_ffn
Add
1 × 163840 × 7168
66added—RMSNorm_11_1
Rms Norm
7.2K1 × 163840 × 7168
67added—MLA_11
Mla
110M1 × 163840 × 7168
68added—Add_11_attn
Add
1 × 163840 × 7168
69added—RMSNorm_11_2
Rms Norm
7.2K1 × 163840 × 7168
70added—MoE_11
Shared Expert Mo E
11B1 × 163840 × 7168
71added—Add_11_ffn
Add
1 × 163840 × 7168
72added—RMSNorm_12_1
Rms Norm
7.2K1 × 163840 × 7168
73added—MLA_12
Mla
110M1 × 163840 × 7168
74added—Add_12_attn
Add
1 × 163840 × 7168
75added—RMSNorm_12_2
Rms Norm
7.2K1 × 163840 × 7168
76added—MoE_12
Shared Expert Mo E
11B1 × 163840 × 7168
77added—Add_12_ffn
Add
1 × 163840 × 7168
78added—RMSNorm_13_1
Rms Norm
7.2K1 × 163840 × 7168
79added—MLA_13
Mla
110M1 × 163840 × 7168
80added—Add_13_attn
Add
1 × 163840 × 7168
81added—RMSNorm_13_2
Rms Norm
7.2K1 × 163840 × 7168
82added—MoE_13
Shared Expert Mo E
11B1 × 163840 × 7168
83added—Add_13_ffn
Add
1 × 163840 × 7168
84added—RMSNorm_14_1
Rms Norm
7.2K1 × 163840 × 7168
85added—MLA_14
Mla
110M1 × 163840 × 7168
86added—Add_14_attn
Add
1 × 163840 × 7168
87added—RMSNorm_14_2
Rms Norm
7.2K1 × 163840 × 7168
88added—MoE_14
Shared Expert Mo E
11B1 × 163840 × 7168
89added—Add_14_ffn
Add
1 × 163840 × 7168
90added—RMSNorm_15_1
Rms Norm
7.2K1 × 163840 × 7168
91added—MLA_15
Mla
110M1 × 163840 × 7168
92added—Add_15_attn
Add
1 × 163840 × 7168
93added—RMSNorm_15_2
Rms Norm
7.2K1 × 163840 × 7168
94added—MoE_15
Shared Expert Mo E
11B1 × 163840 × 7168
95added—Add_15_ffn
Add
1 × 163840 × 7168
96added—RMSNorm_16_1
Rms Norm
7.2K1 × 163840 × 7168
97added—MLA_16
Mla
110M1 × 163840 × 7168
98added—Add_16_attn
Add
1 × 163840 × 7168
99added—RMSNorm_16_2
Rms Norm
7.2K1 × 163840 × 7168
100added—MoE_16
Shared Expert Mo E
11B1 × 163840 × 7168
101added—Add_16_ffn
Add
1 × 163840 × 7168
102added—RMSNorm_17_1
Rms Norm
7.2K1 × 163840 × 7168
103added—MLA_17
Mla
110M1 × 163840 × 7168
104added—Add_17_attn
Add
1 × 163840 × 7168
105added—RMSNorm_17_2
Rms Norm
7.2K1 × 163840 × 7168
106added—MoE_17
Shared Expert Mo E
11B1 × 163840 × 7168
107added—Add_17_ffn
Add
1 × 163840 × 7168
108added—RMSNorm_18_1
Rms Norm
7.2K1 × 163840 × 7168
109added—MLA_18
Mla
110M1 × 163840 × 7168
110added—Add_18_attn
Add
1 × 163840 × 7168
111added—RMSNorm_18_2
Rms Norm
7.2K1 × 163840 × 7168
112added—MoE_18
Shared Expert Mo E
11B1 × 163840 × 7168
113added—Add_18_ffn
Add
1 × 163840 × 7168
114added—RMSNorm_19_1
Rms Norm
7.2K1 × 163840 × 7168
115added—MLA_19
Mla
110M1 × 163840 × 7168
116added—Add_19_attn
Add
1 × 163840 × 7168
117added—RMSNorm_19_2
Rms Norm
7.2K1 × 163840 × 7168
118added—MoE_19
Shared Expert Mo E
11B1 × 163840 × 7168
119added—Add_19_ffn
Add
1 × 163840 × 7168
120added—RMSNorm_20_1
Rms Norm
7.2K1 × 163840 × 7168
121added—MLA_20
Mla
110M1 × 163840 × 7168
122added—Add_20_attn
Add
1 × 163840 × 7168
123added—RMSNorm_20_2
Rms Norm
7.2K1 × 163840 × 7168
124added—MoE_20
Shared Expert Mo E
11B1 × 163840 × 7168
125added—Add_20_ffn
Add
1 × 163840 × 7168
126added—RMSNorm_21_1
Rms Norm
7.2K1 × 163840 × 7168
127added—MLA_21
Mla
110M1 × 163840 × 7168
128added—Add_21_attn
Add
1 × 163840 × 7168
129added—RMSNorm_21_2
Rms Norm
7.2K1 × 163840 × 7168
130added—MoE_21
Shared Expert Mo E
11B1 × 163840 × 7168
131added—Add_21_ffn
Add
1 × 163840 × 7168
132added—RMSNorm_22_1
Rms Norm
7.2K1 × 163840 × 7168
133added—MLA_22
Mla
110M1 × 163840 × 7168
134added—Add_22_attn
Add
1 × 163840 × 7168
135added—RMSNorm_22_2
Rms Norm
7.2K1 × 163840 × 7168
136added—MoE_22
Shared Expert Mo E
11B1 × 163840 × 7168
137added—Add_22_ffn
Add
1 × 163840 × 7168
138added—RMSNorm_23_1
Rms Norm
7.2K1 × 163840 × 7168
139added—MLA_23
Mla
110M1 × 163840 × 7168
140added—Add_23_attn
Add
1 × 163840 × 7168
141added—RMSNorm_23_2
Rms Norm
7.2K1 × 163840 × 7168
142added—MoE_23
Shared Expert Mo E
11B1 × 163840 × 7168
143added—Add_23_ffn
Add
1 × 163840 × 7168
144added—RMSNorm_24_1
Rms Norm
7.2K1 × 163840 × 7168
145added—MLA_24
Mla
110M1 × 163840 × 7168
146added—Add_24_attn
Add
1 × 163840 × 7168
147added—RMSNorm_24_2
Rms Norm
7.2K1 × 163840 × 7168
148added—MoE_24
Shared Expert Mo E
11B1 × 163840 × 7168
149added—Add_24_ffn
Add
1 × 163840 × 7168
150added—RMSNorm_25_1
Rms Norm
7.2K1 × 163840 × 7168
151added—MLA_25
Mla
110M1 × 163840 × 7168
152added—Add_25_attn
Add
1 × 163840 × 7168
153added—RMSNorm_25_2
Rms Norm
7.2K1 × 163840 × 7168
154added—MoE_25
Shared Expert Mo E
11B1 × 163840 × 7168
155added—Add_25_ffn
Add
1 × 163840 × 7168
156added—RMSNorm_26_1
Rms Norm
7.2K1 × 163840 × 7168
157added—MLA_26
Mla
110M1 × 163840 × 7168
158added—Add_26_attn
Add
1 × 163840 × 7168
159added—RMSNorm_26_2
Rms Norm
7.2K1 × 163840 × 7168
160added—MoE_26
Shared Expert Mo E
11B1 × 163840 × 7168
161added—Add_26_ffn
Add
1 × 163840 × 7168
162added—RMSNorm_27_1
Rms Norm
7.2K1 × 163840 × 7168
163added—MLA_27
Mla
110M1 × 163840 × 7168
164added—Add_27_attn
Add
1 × 163840 × 7168
165added—RMSNorm_27_2
Rms Norm
7.2K1 × 163840 × 7168
166added—MoE_27
Shared Expert Mo E
11B1 × 163840 × 7168
167added—Add_27_ffn
Add
1 × 163840 × 7168
168added—RMSNorm_28_1
Rms Norm
7.2K1 × 163840 × 7168
169added—MLA_28
Mla
110M1 × 163840 × 7168
170added—Add_28_attn
Add
1 × 163840 × 7168
171added—RMSNorm_28_2
Rms Norm
7.2K1 × 163840 × 7168
172added—MoE_28
Shared Expert Mo E
11B1 × 163840 × 7168
173added—Add_28_ffn
Add
1 × 163840 × 7168
174added—RMSNorm_29_1
Rms Norm
7.2K1 × 163840 × 7168
175added—MLA_29
Mla
110M1 × 163840 × 7168
176added—Add_29_attn
Add
1 × 163840 × 7168
177added—RMSNorm_29_2
Rms Norm
7.2K1 × 163840 × 7168
178added—MoE_29
Shared Expert Mo E
11B1 × 163840 × 7168
179added—Add_29_ffn
Add
1 × 163840 × 7168
180added—RMSNorm_30_1
Rms Norm
7.2K1 × 163840 × 7168
181added—MLA_30
Mla
110M1 × 163840 × 7168
182added—Add_30_attn
Add
1 × 163840 × 7168
183added—RMSNorm_30_2
Rms Norm
7.2K1 × 163840 × 7168
184added—MoE_30
Shared Expert Mo E
11B1 × 163840 × 7168
185added—Add_30_ffn
Add
1 × 163840 × 7168
186added—RMSNorm_31_1
Rms Norm
7.2K1 × 163840 × 7168
187added—MLA_31
Mla
110M1 × 163840 × 7168
188added—Add_31_attn
Add
1 × 163840 × 7168
189added—RMSNorm_31_2
Rms Norm
7.2K1 × 163840 × 7168
190added—MoE_31
Shared Expert Mo E
11B1 × 163840 × 7168
191added—Add_31_ffn
Add
1 × 163840 × 7168
192added—RMSNorm_32_1
Rms Norm
7.2K1 × 163840 × 7168
193added—MLA_32
Mla
110M1 × 163840 × 7168
194added—Add_32_attn
Add
1 × 163840 × 7168
195added—RMSNorm_32_2
Rms Norm
7.2K1 × 163840 × 7168
196added—MoE_32
Shared Expert Mo E
11B1 × 163840 × 7168
197added—Add_32_ffn
Add
1 × 163840 × 7168
198added—RMSNorm_33_1
Rms Norm
7.2K1 × 163840 × 7168
199added—MLA_33
Mla
110M1 × 163840 × 7168
200added—Add_33_attn
Add
1 × 163840 × 7168
201added—RMSNorm_33_2
Rms Norm
7.2K1 × 163840 × 7168
202added—MoE_33
Shared Expert Mo E
11B1 × 163840 × 7168
203added—Add_33_ffn
Add
1 × 163840 × 7168
204added—RMSNorm_34_1
Rms Norm
7.2K1 × 163840 × 7168
205added—MLA_34
Mla
110M1 × 163840 × 7168
206added—Add_34_attn
Add
1 × 163840 × 7168
207added—RMSNorm_34_2
Rms Norm
7.2K1 × 163840 × 7168
208added—MoE_34
Shared Expert Mo E
11B1 × 163840 × 7168
209added—Add_34_ffn
Add
1 × 163840 × 7168
210added—RMSNorm_35_1
Rms Norm
7.2K1 × 163840 × 7168
211added—MLA_35
Mla
110M1 × 163840 × 7168
212added—Add_35_attn
Add
1 × 163840 × 7168
213added—RMSNorm_35_2
Rms Norm
7.2K1 × 163840 × 7168
214added—MoE_35
Shared Expert Mo E
11B1 × 163840 × 7168
215added—Add_35_ffn
Add
1 × 163840 × 7168
216added—RMSNorm_36_1
Rms Norm
7.2K1 × 163840 × 7168
217added—MLA_36
Mla
110M1 × 163840 × 7168
218added—Add_36_attn
Add
1 × 163840 × 7168
219added—RMSNorm_36_2
Rms Norm
7.2K1 × 163840 × 7168
220added—MoE_36
Shared Expert Mo E
11B1 × 163840 × 7168
221added—Add_36_ffn
Add
1 × 163840 × 7168
222added—RMSNorm_37_1
Rms Norm
7.2K1 × 163840 × 7168
223added—MLA_37
Mla
110M1 × 163840 × 7168
224added—Add_37_attn
Add
1 × 163840 × 7168
225added—RMSNorm_37_2
Rms Norm
7.2K1 × 163840 × 7168
226added—MoE_37
Shared Expert Mo E
11B1 × 163840 × 7168
227added—Add_37_ffn
Add
1 × 163840 × 7168
228added—RMSNorm_38_1
Rms Norm
7.2K1 × 163840 × 7168
229added—MLA_38
Mla
110M1 × 163840 × 7168
230added—Add_38_attn
Add
1 × 163840 × 7168
231added—RMSNorm_38_2
Rms Norm
7.2K1 × 163840 × 7168
232added—MoE_38
Shared Expert Mo E
11B1 × 163840 × 7168
233added—Add_38_ffn
Add
1 × 163840 × 7168
234added—RMSNorm_39_1
Rms Norm
7.2K1 × 163840 × 7168
235added—MLA_39
Mla
110M1 × 163840 × 7168
236added—Add_39_attn
Add
1 × 163840 × 7168
237added—RMSNorm_39_2
Rms Norm
7.2K1 × 163840 × 7168
238added—MoE_39
Shared Expert Mo E
11B1 × 163840 × 7168
239added—Add_39_ffn
Add
1 × 163840 × 7168
240added—RMSNorm_40_1
Rms Norm
7.2K1 × 163840 × 7168
241added—MLA_40
Mla
110M1 × 163840 × 7168
242added—Add_40_attn
Add
1 × 163840 × 7168
243added—RMSNorm_40_2
Rms Norm
7.2K1 × 163840 × 7168
244added—MoE_40
Shared Expert Mo E
11B1 × 163840 × 7168
245added—Add_40_ffn
Add
1 × 163840 × 7168
246added—RMSNorm_41_1
Rms Norm
7.2K1 × 163840 × 7168
247added—MLA_41
Mla
110M1 × 163840 × 7168
248added—Add_41_attn
Add
1 × 163840 × 7168
249added—RMSNorm_41_2
Rms Norm
7.2K1 × 163840 × 7168
250added—MoE_41
Shared Expert Mo E
11B1 × 163840 × 7168
251added—Add_41_ffn
Add
1 × 163840 × 7168
252added—RMSNorm_42_1
Rms Norm
7.2K1 × 163840 × 7168
253added—MLA_42
Mla
110M1 × 163840 × 7168
254added—Add_42_attn
Add
1 × 163840 × 7168
255added—RMSNorm_42_2
Rms Norm
7.2K1 × 163840 × 7168
256added—MoE_42
Shared Expert Mo E
11B1 × 163840 × 7168
257added—Add_42_ffn
Add
1 × 163840 × 7168
258added—RMSNorm_43_1
Rms Norm
7.2K1 × 163840 × 7168
259added—MLA_43
Mla
110M1 × 163840 × 7168
260added—Add_43_attn
Add
1 × 163840 × 7168
261added—RMSNorm_43_2
Rms Norm
7.2K1 × 163840 × 7168
262added—MoE_43
Shared Expert Mo E
11B1 × 163840 × 7168
263added—Add_43_ffn
Add
1 × 163840 × 7168
264added—RMSNorm_44_1
Rms Norm
7.2K1 × 163840 × 7168
265added—MLA_44
Mla
110M1 × 163840 × 7168
266added—Add_44_attn
Add
1 × 163840 × 7168
267added—RMSNorm_44_2
Rms Norm
7.2K1 × 163840 × 7168
268added—MoE_44
Shared Expert Mo E
11B1 × 163840 × 7168
269added—Add_44_ffn
Add
1 × 163840 × 7168
270added—RMSNorm_45_1
Rms Norm
7.2K1 × 163840 × 7168
271added—MLA_45
Mla
110M1 × 163840 × 7168
272added—Add_45_attn
Add
1 × 163840 × 7168
273added—RMSNorm_45_2
Rms Norm
7.2K1 × 163840 × 7168
274added—MoE_45
Shared Expert Mo E
11B1 × 163840 × 7168
275added—Add_45_ffn
Add
1 × 163840 × 7168
276added—RMSNorm_46_1
Rms Norm
7.2K1 × 163840 × 7168
277added—MLA_46
Mla
110M1 × 163840 × 7168
278added—Add_46_attn
Add
1 × 163840 × 7168
279added—RMSNorm_46_2
Rms Norm
7.2K1 × 163840 × 7168
280added—MoE_46
Shared Expert Mo E
11B1 × 163840 × 7168
281added—Add_46_ffn
Add
1 × 163840 × 7168
282added—RMSNorm_47_1
Rms Norm
7.2K1 × 163840 × 7168
283added—MLA_47
Mla
110M1 × 163840 × 7168
284added—Add_47_attn
Add
1 × 163840 × 7168
285added—RMSNorm_47_2
Rms Norm
7.2K1 × 163840 × 7168
286added—MoE_47
Shared Expert Mo E
11B1 × 163840 × 7168
287added—Add_47_ffn
Add
1 × 163840 × 7168
288added—RMSNorm_48_1
Rms Norm
7.2K1 × 163840 × 7168
289added—MLA_48
Mla
110M1 × 163840 × 7168
290added—Add_48_attn
Add
1 × 163840 × 7168
291added—RMSNorm_48_2
Rms Norm
7.2K1 × 163840 × 7168
292added—MoE_48
Shared Expert Mo E
11B1 × 163840 × 7168
293added—Add_48_ffn
Add
1 × 163840 × 7168
294added—RMSNorm_49_1
Rms Norm
7.2K1 × 163840 × 7168
295added—MLA_49
Mla
110M1 × 163840 × 7168
296added—Add_49_attn
Add
1 × 163840 × 7168
297added—RMSNorm_49_2
Rms Norm
7.2K1 × 163840 × 7168
298added—MoE_49
Shared Expert Mo E
11B1 × 163840 × 7168
299added—Add_49_ffn
Add
1 × 163840 × 7168
300added—RMSNorm_50_1
Rms Norm
7.2K1 × 163840 × 7168
301added—MLA_50
Mla
110M1 × 163840 × 7168
302added—Add_50_attn
Add
1 × 163840 × 7168
303added—RMSNorm_50_2
Rms Norm
7.2K1 × 163840 × 7168
304added—MoE_50
Shared Expert Mo E
11B1 × 163840 × 7168
305added—Add_50_ffn
Add
1 × 163840 × 7168
306added—RMSNorm_51_1
Rms Norm
7.2K1 × 163840 × 7168
307added—MLA_51
Mla
110M1 × 163840 × 7168
308added—Add_51_attn
Add
1 × 163840 × 7168
309added—RMSNorm_51_2
Rms Norm
7.2K1 × 163840 × 7168
310added—MoE_51
Shared Expert Mo E
11B1 × 163840 × 7168
311added—Add_51_ffn
Add
1 × 163840 × 7168
312added—RMSNorm_52_1
Rms Norm
7.2K1 × 163840 × 7168
313added—MLA_52
Mla
110M1 × 163840 × 7168
314added—Add_52_attn
Add
1 × 163840 × 7168
315added—RMSNorm_52_2
Rms Norm
7.2K1 × 163840 × 7168
316added—MoE_52
Shared Expert Mo E
11B1 × 163840 × 7168
317added—Add_52_ffn
Add
1 × 163840 × 7168
318added—RMSNorm_53_1
Rms Norm
7.2K1 × 163840 × 7168
319added—MLA_53
Mla
110M1 × 163840 × 7168
320added—Add_53_attn
Add
1 × 163840 × 7168
321added—RMSNorm_53_2
Rms Norm
7.2K1 × 163840 × 7168
322added—MoE_53
Shared Expert Mo E
11B1 × 163840 × 7168
323added—Add_53_ffn
Add
1 × 163840 × 7168
324added—RMSNorm_54_1
Rms Norm
7.2K1 × 163840 × 7168
325added—MLA_54
Mla
110M1 × 163840 × 7168
326added—Add_54_attn
Add
1 × 163840 × 7168
327added—RMSNorm_54_2
Rms Norm
7.2K1 × 163840 × 7168
328added—MoE_54
Shared Expert Mo E
11B1 × 163840 × 7168
329added—Add_54_ffn
Add
1 × 163840 × 7168
330added—RMSNorm_55_1
Rms Norm
7.2K1 × 163840 × 7168
331added—MLA_55
Mla
110M1 × 163840 × 7168
332added—Add_55_attn
Add
1 × 163840 × 7168
333added—RMSNorm_55_2
Rms Norm
7.2K1 × 163840 × 7168
334added—MoE_55
Shared Expert Mo E
11B1 × 163840 × 7168
335added—Add_55_ffn
Add
1 × 163840 × 7168
336added—RMSNorm_56_1
Rms Norm
7.2K1 × 163840 × 7168
337added—MLA_56
Mla
110M1 × 163840 × 7168
338added—Add_56_attn
Add
1 × 163840 × 7168
339added—RMSNorm_56_2
Rms Norm
7.2K1 × 163840 × 7168
340added—MoE_56
Shared Expert Mo E
11B1 × 163840 × 7168
341added—Add_56_ffn
Add
1 × 163840 × 7168
342added—RMSNorm_57_1
Rms Norm
7.2K1 × 163840 × 7168
343added—MLA_57
Mla
110M1 × 163840 × 7168
344added—Add_57_attn
Add
1 × 163840 × 7168
345added—RMSNorm_57_2
Rms Norm
7.2K1 × 163840 × 7168
346added—MoE_57
Shared Expert Mo E
11B1 × 163840 × 7168
347added—Add_57_ffn
Add
1 × 163840 × 7168
348added—RMSNorm_58_1
Rms Norm
7.2K1 × 163840 × 7168
349added—MLA_58
Mla
110M1 × 163840 × 7168
350added—Add_58_attn
Add
1 × 163840 × 7168
351added—RMSNorm_58_2
Rms Norm
7.2K1 × 163840 × 7168
352added—MoE_58
Shared Expert Mo E
11B1 × 163840 × 7168
353added—Add_58_ffn
Add
1 × 163840 × 7168
354added—RMSNorm_59_1
Rms Norm
7.2K1 × 163840 × 7168
355added—MLA_59
Mla
110M1 × 163840 × 7168
356added—Add_59_attn
Add
1 × 163840 × 7168
357added—RMSNorm_59_2
Rms Norm
7.2K1 × 163840 × 7168
358added—MoE_59
Shared Expert Mo E
11B1 × 163840 × 7168
359added—Add_59_ffn
Add
1 × 163840 × 7168
360added—RMSNorm_60_1
Rms Norm
7.2K1 × 163840 × 7168
361added—MLA_60
Mla
110M1 × 163840 × 7168
362added—Add_60_attn
Add
1 × 163840 × 7168
363added—RMSNorm_60_2
Rms Norm
7.2K1 × 163840 × 7168
364added—MoE_60
Shared Expert Mo E
11B1 × 163840 × 7168
365added—Add_60_ffn
Add
1 × 163840 × 7168
366added—RMSNorm_61_1
Rms Norm
7.2K1 × 163840 × 7168
367added—MLA_61
Mla
110M1 × 163840 × 7168
368added—Add_61_attn
Add
1 × 163840 × 7168
369added—RMSNorm_61_2
Rms Norm
7.2K1 × 163840 × 7168
370added—MoE_61
Shared Expert Mo E
11B1 × 163840 × 7168
371sameresidual_2
Add
1 × 2048 × 4096Add_61_ffn
Add
1 × 163840 × 7168
372added—Final_RMSNorm
Rms Norm
7.2K1 × 163840 × 7168
373added—LM_Head
Linear
927M1 × 163840 × 129280
374samehidden_state
Output
1 × 2048 × 4096Output
Output
1 × 163840 × 129280

What this is not

Take it further

Open either graph in the editor, change it, and check it again: LLaMA-3 Block · DeepSeek-V3

Compare any two models of your own, including anything on Hugging Face: the comparison tool.

Machine-readable: this page as markdown · the pair index · POST https://www.neurarch.com/api/v1/plan for a graph of your own.

Related comparisons

LLaMA-3 Block vs Mixtral MoE Block
A dense feed-forward against a mixture of experts, at the block level.
Mixtral MoE Block vs DeepSeek-V3
Two mixture-of-experts designs that route very differently.
Transformer Block vs LLaMA-3 Block
What changed in the transformer block between 2017 and now.
LLaMA-3 Block vs Phi-3 Mini Block
A block from a large model against a whole small one.