DeepSeek V3.1 Terminus deepseek-v3.1-terminus

deepseek-ai · deepseek hybrid-reasoning

unranked — no quality signals yet

params
671B (A37B)
arch
moe
context
160k
license
mit open weights
released
Sep 2025
reasoning
yes
downloads/30d
10.6k

# why ranked

Overall: provisional — score 26.3 ● low — a single signal; treat with caution

provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.

signalweightinput (0–100)
aa_intelligence 0.50 26.3
arena_elo 0.30
bench_composite 0.20

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

Coding: provisional — score 48.3 ● low — a single signal; treat with caution

provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.

signalweightinput (0–100)
aa_coding 0.25 48.3
aider_polyglot 0.25
arena_elo_code 0.15
swe_bench_verified 0.35

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

full methodology

# trend

collecting history — trends appear after a few daily runs.

# benchmarks

benchmarkscoresourcedate
AA Coding Index 43.5 Artificial Analysis
AA Intelligence Index 26.3 Artificial Analysis
AA Math Index 89.7 Artificial Analysis

# where to run

providerquantctx$/M in$/M out$/M blendtpsuptime
DeepInfrafp4 160k $0.27$0.95$0.44 99.9%
Novitafp8 128k $0.27$1.00$0.45 100.0%
SiliconFlowfp8 160k $0.27$1.00$0.45 99.9%
AtlasCloudfp8 128k $0.30$0.95$0.46 100.0%
StreamLakeunknown 125k $0.34$1.03$0.51 100.0%

blended = (3·input + output) / 4 per 1M tokens · cheapest first

source aliases
openrouter
deepseek/deepseek-v3.1-terminus