Skip to content
fedi.software

Step 3.5 Flash in fedi-index

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

fedi-index · General

B 1,381 ±2

Percentile 59 · #121 · provisional

People prefer / Solves tasks

1,392 / —

Human votes vs benchmarks, Elo-eq

Value

+50

$0.15 per 1M tokens, median of 11 providers · vendor $0.15

On your hardware

  • GPU 12 GB #25 · IQ2_S · strong compression · ≈15 tok/s
  • GPU 24 GB #28 · Q2_K_L · strong compression · ≈16 tok/s
  • Mac 64 GB #19 · IQ1_M · strong compression · ≈45 tok/s
  • 128 GB unified memory #8 · IQ4_XS · ≈17 tok/s

Score decomposition

B General 1,381 [1,380–1,383] Percentile 59 · #121 · components: 6 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/overall 1,403 ±1.80 1,403 ±2 25.6% 2026-10-02 —
LMArena text/hard_prompts 1,411 ±2.18 1,405 ±2 25.5% 2026-10-02 —
LMArena text/instruction_following 1,385 ±2.67 1,399 ±2 12.6% 2026-10-02 —
LMArena text/expert 1,422 ±4.31 1,412 ±4 12.3% 2026-10-02 —
LMArena text/longer_query 1,402 ±2.62 1,403 ±2 6.3% 2026-10-02 —
LMArena text/multi_turn 1,405 ±3.31 1,404 ±3 6.2% 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 1,393 [1,388–1,398] Percentile 64 · #100 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/coding 1,436 ±2.89 1,413 ±3 90.3% 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

B Math 1,377 [1,368–1,386] Percentile 55 · #116 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/math 1,408 ±5.21 1,401 ±5 87.8% 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 Writing 1,379 [1,375–1,384] Percentile 57 · #116 · components: 2 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/creative_writing 1,357 ±3.43 1,385 ±4 62.0% 2026-10-02 —
LMArena text/longer_query 1,402 ±2.62 1,403 ±2 31.8% 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: 繁體中文 1,398 [1,390–1,405] Percentile 55 · #101 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/chinese 1,447 ±5.13 1,411 ±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 1,391 [1,376–1,406] Percentile 51 · #71 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/german 1,404 ±9.58 1,406 ±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: English 1,394 [1,389–1,399] Percentile 61 · #108 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/english 1,417 ±2.42 1,406 ±3 94.0% 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 Language: Español 1,402 [1,390–1,413] Percentile 46 · #81 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/spanish 1,419 ±7.09 1,416 ±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: Français 1,392 [1,380–1,405] Percentile 40 · #81 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/french 1,421 ±7.49 1,406 ±7 93.0% 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 Language: 日本語 1,376 [1,360–1,393] Percentile 41 · #76 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/japanese 1,353 ±12 1,391 ±9 92.1% 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: 한국어 1,384 [1,370–1,397] Percentile 40 · #82 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/korean 1,351 ±9.65 1,398 ±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: Polski 1,384 [1,371–1,397] Percentile 34 · #87 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/polish 1,389 ±8.27 1,398 ±7 92.9% 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

B Language: Русский 1,382 [1,376–1,388] Percentile 50 · #117 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/russian 1,385 ±3.70 1,393 ±3 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
Step 3.5 Flash 1,381 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: Step 3.5 Flash

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