Skip to content
fedi.software

DeepSeek V3.2 in fedi-index

The score in every slice, what it is made of, value, local run and history.

fedi-index · General

B 1,403 ±5

Percentile 70 · #90

People prefer / Solves tasks

1,411 / 1,394

Human votes vs benchmarks, Elo-eq

Value

+54

$0.315 per 1M tokens, median of 30 providers

On your hardware

Closed weights: cloud only.

Score decomposition

B General 1,403 [1,398–1,409] Percentile 70 · #90 · components: 7
Source Component Value Elo-eq Weight Measured Variant
Epoch AI eci 146 ±0.76 1,408 ±6 44.0% 2025-12-01 —
LMArena text/overall 1,424 ±1.77 1,424 ±2 14.4% 2026-10-02 —
LMArena text/hard_prompts 1,433 ±2.26 1,425 ±2 14.3% 2026-10-02 —
LMArena text/instruction_following 1,412 ±2.91 1,425 ±3 7.1% 2026-10-02 —
LMArena text/expert 1,436 ±5.23 1,424 ±4 6.8% 2026-10-02 —
LMArena text/longer_query 1,428 ±2.92 1,426 ±3 3.5% 2026-10-02 —
LMArena text/multi_turn 1,427 ±3.60 1,424 ±3 3.5% 2026-10-02 —

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Coding 1,399 [1,396–1,402] Percentile 67 · #92 · components: 2 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena webdev/overall 1,362 ±4.62 1,410 ±1 45.9% 2026-10-01 thinking
LMArena text/coding 1,454 ±3.39 1,429 ±3 44.5% 2026-10-02 thinking

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Math 1,399 [1,390–1,408] Percentile 70 · #77 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/math 1,435 ±5.42 1,427 ±5 87.7% 2026-10-02 —

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Writing 1,414 [1,408–1,419] Percentile 71 · #78 · components: 2 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/creative_writing 1,400 ±3.86 1,428 ±4 61.8% 2026-10-02 —
LMArena text/longer_query 1,428 ±2.92 1,426 ±3 32.0% 2026-10-02 —

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Language: 繁體中文 1,405 [1,397–1,412] Percentile 60 · #90 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/chinese 1,456 ±5.14 1,418 ±4 93.7% 2026-10-02 —

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Language: Deutsch 1,397 [1,380–1,414] Percentile 53 · #68 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/german 1,413 ±11 1,414 ±9 92.0% 2026-10-02 thinking

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Language: English 1,416 [1,411–1,421] Percentile 70 · #82 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/english 1,439 ±2.60 1,430 ±3 93.9% 2026-10-02 thinking

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Language: Español 1,404 [1,391–1,418] Percentile 51 · #73 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/spanish 1,423 ±8.03 1,420 ±7 92.8% 2026-10-02 —

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

C Language: Français 1,397 [1,382–1,412] Percentile 45 · #75 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/french 1,429 ±9.06 1,413 ±8 92.4% 2026-10-02 —

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

C Language: 日本語 1,391 [1,371–1,411] Percentile 46 · #70 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/japanese 1,377 ±15 1,409 ±11 91.1% 2026-10-02 thinking

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Language: 한국어 1,397 [1,382–1,412] Percentile 50 · #68 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/korean 1,371 ±11 1,413 ±8 92.5% 2026-10-02 —

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Language: Polski 1,408 [1,394–1,421] Percentile 52 · #64 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/polish 1,419 ±8.74 1,424 ±7 92.8% 2026-10-02 thinking

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

B Language: Русский 1,414 [1,408–1,421] Percentile 69 · #74 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/russian 1,424 ±3.94 1,428 ±4 93.8% 2026-10-02 —

Weight is the share of the component in the score (the rest is the prior that pulls models with little data towards the population median). Methodology

History

fedi-index by month
The same in a table
Model 2026-10
DeepSeek V3.2 1,403 B

Badge

Show the model's fedi-index on your site or in a README. The badge updates itself; it links to this page.

fedi-index: DeepSeek V3.2

Data CC BY 4.0: keep the link to fedi.software.