Gemini 3 Flash Preview in fedi-index
The score in every slice, what it is made of, value, local run and history.
fedi-index · General
A 1,437 ±4
Percentile 86 · #43
People prefer / Solves tasks
1,440 / 1,434
Human votes vs benchmarks, Elo-eq
Value
+57
$1.12 per 1M tokens, median of 26 providers · vendor $1.12
On your hardware
Closed weights: cloud only.
Score decomposition
A General 1,437 [1,433–1,442] Percentile 86 · #43 · components: 9
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| Epoch AI |
eci
|
152 ±0.75 | 1,452 ±6 | 36.3% | 2025-12-17 | — |
| LMArena |
text/overall
|
1,466 ±2.21 | 1,466 ±2 | 15.6% | 2026-10-02 | — |
| LMArena |
text/hard_prompts
|
1,464 ±2.81 | 1,453 ±2 | 15.6% | 2026-10-02 | — |
| LMArena |
text/instruction_following
|
1,436 ±3.63 | 1,448 ±3 | 7.6% | 2026-10-02 | — |
| LMArena |
text/expert
|
1,462 ±6.35 | 1,445 ±5 | 7.1% | 2026-10-02 | — |
| LMArena |
text/longer_query
|
1,452 ±3.63 | 1,449 ±3 | 3.8% | 2026-10-02 | — |
| LMArena |
text/multi_turn
|
1,471 ±4.33 | 1,463 ±4 | 3.7% | 2026-10-02 | — |
| Epoch AI |
simpleqa_verified
outlier decay ×0.61 |
0.67 ±0.01 | 1,507 ±5 | 1.9% | 2026-08-27 | high |
| Epoch AI |
gpqa_diamond
decay ×0.26 |
0.89 ±0.02 | 1,456 ±10 | 1.3% | 2026-08-06 | 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
B Coding 1,418 [1,414–1,422] Percentile 76 · #66 · components: 3
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
webdev/overall
|
1,439 ±4.38 | 1,432 ±1 | 39.9% | 2026-10-01 | — |
| LMArena |
text/coding
|
1,460 ±3.95 | 1,434 ±4 | 38.1% | 2026-10-02 | — |
| Epoch AI |
swe_bench_verified
decay ×0.37 |
0.75 ±0.02 | 1,460 ±10 | 13.7% | 2026-02-18 | — |
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 Agents 1,406 [1,395–1,418] Percentile 25 · #44 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| τ²-bench (Sierra) |
tau2/banking_knowledge
decay ×0.76 |
0.27 ±0.05 | 1,459 ±8 | 79.5% | 2026-03-02 | 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 Math 1,429 [1,421–1,437] Percentile 85 · #39 · components: 3
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/math
|
1,472 ±6.57 | 1,463 ±6 | 51.0% | 2026-10-02 | — |
| Epoch AI |
frontiermath_t1_3
decay ×0.86 |
0.51 ±0.03 | 1,432 ±6 | 29.2% | 2026-06-11 | — |
| Epoch AI |
frontiermath_t4
decay ×0.86 |
0.17 ±0.06 | 1,420 ±10 | 12.3% | 2026-06-11 | — |
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 1,445 [1,442–1,448] Percentile 80 · #24 · components: 3 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
vision/overall
|
1,285 ±2.67 | 1,462 ±2 | 62.8% | 2026-10-02 | — |
| LMArena |
vision/ocr
|
1,291 ±2.59 | 1,463 ±2 | 15.7% | 2026-10-02 | — |
| LMArena |
vision/diagram
|
1,290 ±3.59 | 1,454 ±3 | 15.4% | 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 Writing 1,456 [1,450–1,462] Percentile 90 · #28 · components: 2 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/creative_writing
|
1,457 ±4.70 | 1,486 ±5 | 61.4% | 2026-10-02 | — |
| LMArena |
text/longer_query
|
1,452 ±3.63 | 1,449 ±3 | 32.2% | 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
A Language: 繁體中文 1,445 [1,436–1,454] Percentile 86 · #32 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/chinese
|
1,510 ±6.15 | 1,461 ±5 | 93.5% | 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
A Language: Deutsch 1,461 [1,441–1,480] Percentile 96 · #6 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/german
|
1,497 ±13 | 1,486 ±11 | 91.2% | 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
A Language: English 1,443 [1,437–1,449] Percentile 87 · #36 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/english
|
1,467 ±3.16 | 1,459 ±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
A Language: Español 1,443 [1,427–1,459] Percentile 87 · #20 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/spanish
|
1,470 ±9.60 | 1,463 ±9 | 92.2% | 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 Language: Français 1,437 [1,419–1,455] Percentile 80 · #28 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/french
|
1,478 ±11 | 1,458 ±10 | 91.7% | 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: 日本語 1,464 [1,441–1,487] Percentile 91 · #12 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/japanese
|
1,488 ±17 | 1,492 ±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
A Language: 한국어 1,446 [1,429–1,464] Percentile 89 · #16 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/korean
|
1,443 ±13 | 1,468 ±10 | 91.8% | 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
A Language: Polski 1,466 [1,451–1,481] Percentile 95 · #8 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/polish
|
1,494 ±9.72 | 1,487 ±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
A Language: Русский 1,462 [1,454–1,469] Percentile 93 · #18 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/russian
|
1,480 ±4.72 | 1,479 ±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
History
The same in a table
| Model | 2026-10 |
|---|---|
| Gemini 3 Flash Preview | 1,437 A |
Badge
Show the model's fedi-index on your site or in a README. The badge updates itself; it links to this page.
Data CC BY 4.0: keep the link to fedi.software.