DeepSeek V3 deepseek-v3

deepseek-ai · deepseek flagship

#42 overall #5 coding #19 agentic

params
671B (A37B)
arch
moe
context
160k
license
mit open weights
released
Dec 2024
train compute
3.3e+24 FLOP
downloads/30d
1.0M

# architecture

attn shared MHA ×128 router top-8 of 256 +1 shared expert ×256 expert out ×61 layers ctx 163,840 671B pool · A37B/token
layers
61
d_model
7168
heads
128q / 128kv
head dim
56
experts
top-8 of 256 + 1 shared
vocab
129280
family
deepseek_v3

signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.

# why ranked

Overall: #42 — score 33.8 ● high — all signals present

signalweightinput (0–100)
aa_intelligence 0.70 33.5
bench_composite 0.30 34.7

benchmark panel evidence:

complete panelscore (0–100)
aider-polyglot-v141.8
swe-bench-verified-v127.5

Coding: #5 — score 28.2 ● high — all signals present

signalweightinput (0–100)
aa_coding 0.30 24.6
aider_polyglot 0.30 54.5
swe_bench_verified 0.40 11.1

Agentic: #19 — score 11.1 ● low — a single signal; treat with caution

signalweightinput (0–100)
swe_bench_verified 0.60 11.1
swe_rebench 0.40

missing signals are dropped and the remaining weights renormalized — never imputed.

full methodology

# trend

OGM score · last 55 days
OGM score over 55 days
overall rank (up is better)
Overall rank over 55 days

methodology changed during this history window; score movement across that boundary is not model movement. See methodology v5.

# benchmarks

benchmarkscoresourcedate
AA Coding Index 23.0 Artificial Analysis
AA Intelligence Index 9.2 Artificial Analysis
AA Math Index 41.0 Artificial Analysis
Aider Polyglot 55.1 Aider Mar 2025
Aider-polyglot 0.5 / 1 LLM Stats
Aider-polyglot-edit 0.8 / 1 LLM Stats
AIME 2024 0.6 / 1 LLM Stats
C-eval 0.9 / 1 LLM Stats
Cluewsc 0.9 / 1 LLM Stats
Cnmo-2024 0.4 / 1 LLM Stats
Csimpleqa 0.6 / 1 LLM Stats
SWE-bench Verified 42.0 SWE-bench Aug 2025

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime
SiliconFlow siliconflow
Hyperbolic 32k $0.20$0.20 via litellm
DeepSeek 128k $0.28$0.42 via litellm
StreamLakeunknown 125k $0.26$1.03 via openrouter 99.4%
DeepInfrafp4 160k $0.32$0.89 deepinfra 94.0%
Novita 64k $0.40$1.30 novita
Nebius 125k $0.50$1.50 via litellm
Amazon Bedrock 160k $0.58$1.68 via litellm
Together 128k $1.25$1.25 via modelsdev
Replicate 64k $1.45$1.45 via litellm
Azure 125k $1.14$4.56 via litellm

sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price

# variants

DeepSeek R1 DeepSeek V3.1 DeepSeek V3.2

source aliases
aa
deepseek-v3, deepseek-v3-0324
aider
DeepSeek Chat V3 (prev), DeepSeek V3 (0324)
arena
DeepSeek-V3
epoch
DeepSeek-V3 (Mar 2025)
litellm
deepseek/deepseek-chat
llmstats
deepseek-v3-0324
openrouter
deepseek/deepseek-chat
swebench
DeepSeek-V3-0324