Skip to content
fedi.software

Qwen3.8 27B in fedi-index

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

fedi-index · General

A 1,425 ±7

Percentile 81 · #57

People prefer / Solves tasks

1,432 / 1,416

Human votes vs benchmarks, Elo-eq

Value

+46

$1.04 per 1M tokens, median of 39 providers

On your hardware

  • GPU 12 GB #10 · UD-Q4_K_XL · ≈4 tok/s
  • GPU 24 GB #1 · UD-Q5_K_S · ≈38 tok/s
  • Mac 64 GB #1 · Q8_0 · ≈13 tok/s
  • 128 GB unified memory #6 · UD-Q4_K_M · ≈10 tok/s
  • CPU, 32 GB RAM #5 · UD-Q4_K_XL · ≈2 tok/s

Score decomposition

A General 1,425 [1,418–1,431] Percentile 81 · #57 · components: 7
Source Component Value Elo-eq Weight Measured Variant
Epoch AI eci 149 ±0.98 1,433 ±8 42.5% 2026-08-14 —
LMArena text/overall 1,441 ±2.74 1,441 ±3 14.9% 2026-10-02 —
LMArena text/hard_prompts 1,461 ±3.22 1,450 ±3 14.8% 2026-10-02 —
LMArena text/instruction_following 1,438 ±4.04 1,449 ±4 7.2% 2026-10-02 —
LMArena text/expert 1,481 ±6.76 1,461 ±6 6.8% 2026-10-02 —
LMArena text/longer_query 1,448 ±3.79 1,446 ±4 3.6% 2026-10-02 —
LMArena text/multi_turn 1,442 ±5.71 1,437 ±5 3.4% 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

A Coding 1,441 [1,438–1,445] Percentile 91 · #27 · components: 2 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena webdev/overall 1,591 ±3.57 1,476 ±1 46.5% 2026-10-01 —
LMArena text/coding 1,485 ±4.42 1,457 ±4 43.7% 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

C Agents 1,424 [1,422–1,427] Percentile 39 · #36 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena agent/overall -0.02 ±0.00 1,448 ±2 90.5% 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 Math 1,413 [1,397–1,429] Percentile 79 · #55 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/math 1,459 ±9.93 1,450 ±10 85.1% 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 Vision 1,433 [1,428–1,439] Percentile 70 · #35 · components: 3 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena vision/overall 1,270 ±4.41 1,448 ±4 64.0% 2026-10-02 —
LMArena vision/ocr 1,281 ±4.95 1,453 ±5 15.5% 2026-10-02 —
LMArena vision/diagram 1,287 ±7.83 1,452 ±7 14.1% 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 1,408 [1,401–1,414] Percentile 68 · #87 · components: 2 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/creative_writing 1,381 ±5.42 1,409 ±6 60.8% 2026-10-02 —
LMArena text/longer_query 1,448 ±3.79 1,446 ±4 32.7% 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

A Language: 繁體中文 1,440 [1,427–1,452] Percentile 83 · #39 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/chinese 1,505 ±8.30 1,457 ±7 93.1% 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: English 1,422 [1,414–1,430] Percentile 74 · #71 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/english 1,446 ±3.86 1,437 ±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: Español 1,422 [1,401–1,443] Percentile 68 · #49 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/spanish 1,450 ±13 1,444 ±12 90.8% 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

B Language: Français 1,426 [1,405–1,447] Percentile 69 · #42 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/french 1,467 ±13 1,448 ±12 90.8% 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

B Language: 日本語 1,402 [1,379–1,425] Percentile 53 · #61 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/japanese 1,396 ±17 1,424 ±13 90.1% 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: 한국어 1,412 [1,388–1,435] Percentile 63 · #51 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/korean 1,400 ±17 1,435 ±13 90.0% 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: Русский 1,407 [1,397–1,418] Percentile 65 · #82 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/russian 1,417 ±6.38 1,422 ±6 93.3% 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
Qwen3.8 27B 1,425 A

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: Qwen3.8 27B

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