Skip to content
fedi.software

Gemini 4 Carbon in fedi-index

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

fedi-index · General

S 1,524 ±8

Percentile 100 · #2

People prefer / Solves tasks

1,514 / 1,539

Human votes vs benchmarks, Elo-eq

Value

+107◆

$4 per 1M tokens, median of 15 providers · vendor $4

On your hardware

Closed weights: cloud only.

Score decomposition

S General 1,524 [1,515–1,532] Percentile 100 · #2 · components: 9
Source Component Value Elo-eq Weight Measured Variant
Epoch AI eci 170 ±1.79 1,594 ±14 28.1% 2026-10-28 —
LMArena text/overall 1,538 ±4.88 1,538 ±5 18.3% 2026-11-11 high
LMArena text/hard_prompts 1,551 ±6.14 1,531 ±6 17.8% 2026-11-11 high
LMArena text/instruction_following 1,539 ±8.36 1,545 ±8 7.9% 2026-11-11 high
LMArena text/expert 1,582 ±15 1,544 ±12 6.0% 2026-11-11 high
Epoch AI simpleqa_verified decay ×0.58 0.78 ±0.01 1,541 ±4 4.8% 2026-11-02 high
LMArena text/longer_query 1,549 ±7.49 1,539 ±7 4.2% 2026-11-11 high
LMArena text/multi_turn 1,569 ±13 1,552 ±12 3.1% 2026-11-11 high
Epoch AI gpqa_diamond outlier decay ×0.25 0.97 ±0.01 1,489 ±6 1.0% 2026-11-02 high

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

S Coding 1,496 [1,490–1,501] Percentile 99 · #3 · components: 3
Source Component Value Elo-eq Weight Measured Variant
LMArena webdev/overall 1,768 ±7.84 1,526 ±2 36.6% 2026-11-15 high
LMArena text/coding 1,564 ±9.52 1,528 ±8 28.3% 2026-11-11 high
Terminal-Bench tbench/terminal-bench decay ×0.90 0.65 ±0.02 1,523 ±3 25.8% 2026-10-28 high

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

S Agents 1,503 [1,500–1,506] Percentile 97 · #3 · components: 2
Source Component Value Elo-eq Weight Measured Variant
LMArena agent/overall 0.16 ±0.00 1,534 ±2 67.8% 2026-11-15 high
Terminal-Bench tbench/terminal-bench decay ×0.90 0.65 ±0.02 1,523 ±3 23.6% 2026-10-28 high

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 Math 1,480 [1,475–1,485] Percentile 98 · #7 · components: 2 provisional
Source Component Value Elo-eq Weight Measured Variant
Epoch AI frontiermath_t1_3 decay ×0.82 0.91 ±0.02 1,517 ±4 58.4% 2026-11-02 high
Epoch AI frontiermath_t4 decay ×0.82 0.95 ±0.03 1,544 ±5 27.4% 2026-11-02 high

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 Writing 1,527 [1,514–1,539] Percentile 100 · #2 · components: 2 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/creative_writing 1,533 ±10 1,564 ±10 57.2% 2026-11-11 high
LMArena text/longer_query 1,549 ±7.49 1,539 ±7 34.9% 2026-11-11 high

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 1,512 [1,490–1,533] Percentile 99 · #3 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/chinese 1,613 ±15 1,543 ±12 90.7% 2026-11-11 high

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: English 1,501 [1,486–1,517] Percentile 100 · #2 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/english 1,530 ±8.09 1,526 ±8 92.4% 2026-11-11 high

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: Русский 1,520 [1,500–1,540] Percentile 100 · #1 · components: 1 provisional
Source Component Value Elo-eq Weight Measured Variant
LMArena text/russian 1,560 ±12 1,551 ±11 91.2% 2026-11-11 high

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 2026-11
Gemini 4 Carbon 1,524 S

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: Gemini 4 Carbon

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