Skip to content
fedi.software

Qwen3 30B A3B Instruct 2507 in fedi-index

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

fedi-index · General

B 1,356 ±9

Percentile 50 · #147

People prefer / Solves tasks

1,375 / 1,320

Human votes vs benchmarks, Elo-eq

Value

+25

$0.15 per 1M tokens, median of 7 providers

On your hardware

  • GPU 12 GB #7 · Q8_0 · ≈16 tok/s
  • GPU 24 GB #9 · Q5_K_S · ≈147 tok/s
  • Mac 64 GB #8 · Q8_0 · ≈45 tok/s
  • 128 GB unified memory #13 · Q8_0 · ≈33 tok/s
  • CPU, 32 GB RAM #4 · Q4_K_M · ≈6 tok/s

Score decomposition

B General 1,356 [1,348–1,365] Percentile 50 · #147 · components: 8
Source Component Value Elo-eq Weight Measured Variant
Epoch AI eci 137 ±1.93 1,338 ±15 28.1% 2025-07-29 —
LMArena text/overall 1,384 ±2.50 1,384 ±2 18.5% 2026-10-02 —
LMArena text/hard_prompts 1,398 ±3.21 1,393 ±3 18.3% 2026-10-02 —
LMArena text/instruction_following 1,364 ±4.02 1,379 ±4 8.9% 2026-10-02 —
LMArena text/expert 1,395 ±8.58 1,390 ±7 7.8% 2026-10-02 —
LMArena text/longer_query 1,378 ±4.37 1,380 ±4 4.4% 2026-10-02 —
LMArena text/multi_turn 1,378 ±4.86 1,379 ±4 4.4% 2026-10-02 —
Epoch AI gpqa_diamond decay ×0.26 0.56 ±0.03 1,298 ±13 1.4% 2026-08-30 —

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 1,376 [1,369–1,383] Percentile 56 · #123 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/coding 1,416 ±4.51 1,396 ±4 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

C Math 1,362 [1,350–1,375] Percentile 50 · #129 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/math 1,394 ±7.73 1,387 ±7 86.5% 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

C Writing 1,347 [1,340–1,354] Percentile 48 · #139 · components: 2 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/creative_writing 1,318 ±5.61 1,344 ±6 60.9% 2026-10-02 —
LMArena text/longer_query 1,378 ±4.37 1,380 ±4 32.5% 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: 繁體中文 1,387 [1,376–1,398] Percentile 51 · #110 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/chinese 1,433 ±7.39 1,400 ±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

C Language: Deutsch 1,368 [1,348–1,388] Percentile 39 · #88 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/german 1,379 ±13 1,384 ±11 91.0% 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: English 1,367 [1,361–1,374] Percentile 53 · #129 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/english 1,391 ±3.35 1,378 ±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

C Language: Español 1,385 [1,366–1,404] Percentile 38 · #93 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/spanish 1,403 ±11 1,402 ±11 91.4% 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: Français 1,382 [1,356–1,408] Percentile 35 · #88 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/french 1,419 ±16 1,404 ±15 88.9% 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 1,367 [1,352–1,382] Percentile 35 · #83 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/japanese 1,338 ±11 1,381 ±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: 한국어 1,369 [1,352–1,387] Percentile 35 · #89 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/korean 1,333 ±13 1,384 ±10 91.9% 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

D Language: Polski 1,356 [1,344–1,368] Percentile 18 · #107 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/polish 1,353 ±7.76 1,368 ±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

C Language: Русский 1,367 [1,355–1,379] Percentile 46 · #128 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/russian 1,369 ±7.32 1,379 ±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

History

fedi-index by month
The same in a table
Model 2026-10
Qwen3 30B A3B Instruct 2507 1,356 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: Qwen3 30B A3B Instruct 2507

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