Gemini 3.8 Flash in fedi-index
The score in every slice, what it is made of, value, local run and history.
fedi-index · General
S 1,472 ±7
Percentile 98 · #7
People prefer / Solves tasks
1,477 / 1,464
Human votes vs benchmarks, Elo-eq
Value
+84◆
$1.50 per 1M tokens, median of 19 providers · vendor $1.50
On your hardware
Closed weights: cloud only.
Score decomposition
S General 1,472 [1,465–1,479] Percentile 98 · #7 · components: 9
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| Epoch AI |
eci
|
157 ±1.49 | 1,491 ±12 | 28.3% | 2026-09-02 | — |
| LMArena |
text/overall
|
1,497 ±2.57 | 1,497 ±3 | 16.9% | 2026-10-02 | high |
| LMArena |
text/hard_prompts
|
1,507 ±3.10 | 1,491 ±3 | 16.8% | 2026-10-02 | high |
| LMArena |
text/instruction_following
|
1,487 ±3.67 | 1,496 ±4 | 8.3% | 2026-10-02 | high |
| LMArena |
text/expert
|
1,527 ±5.61 | 1,498 ±5 | 8.0% | 2026-10-02 | high |
| Epoch AI |
simpleqa_verified
decay ×0.61 |
0.70 ±0.01 | 1,516 ±5 | 4.2% | 2026-09-02 | high |
| LMArena |
text/longer_query
|
1,506 ±3.48 | 1,499 ±3 | 4.2% | 2026-10-02 | high |
| LMArena |
text/multi_turn
|
1,499 ±5.10 | 1,489 ±5 | 4.0% | 2026-10-02 | high |
| Epoch AI |
gpqa_diamond
decay ×0.26 |
0.95 ±0.01 | 1,483 ±6 | 1.7% | 2026-09-02 | 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 Coding 1,450 [1,446–1,453] Percentile 93 · #20 · components: 2 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
webdev/overall
|
1,583 ±4.02 | 1,473 ±1 | 46.3% | 2026-10-01 | high |
| LMArena |
text/coding
|
1,509 ±4.02 | 1,479 ±4 | 44.0% | 2026-10-02 | 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
B Agents 1,444 [1,440–1,448] Percentile 63 · #22 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
agent/overall
|
0.03 ±0.00 | 1,470 ±2 | 90.4% | 2026-10-02 | 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 Math 1,463 [1,454–1,472] Percentile 96 · #12 · components: 3
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/math
|
1,523 ±8.66 | 1,511 ±8 | 48.4% | 2026-10-02 | high |
| Epoch AI |
frontiermath_t1_3
decay ×0.86 |
0.68 ±0.03 | 1,469 ±6 | 31.4% | 2026-09-02 | high |
| Epoch AI |
frontiermath_t4
decay ×0.86 |
0.22 ±0.07 | 1,428 ±10 | 12.3% | 2026-09-02 | 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 Vision 1,474 [1,467–1,482] Percentile 97 · #4 · components: 3 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
vision/overall
|
1,316 ±5.85 | 1,489 ±5 | 65.1% | 2026-10-02 | high |
| LMArena |
vision/ocr
|
1,333 ±6.88 | 1,504 ±7 | 15.3% | 2026-10-02 | high |
| LMArena |
vision/diagram
|
1,353 ±11 | 1,511 ±10 | 12.8% | 2026-10-02 | 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 1,493 [1,487–1,499] Percentile 98 · #7 · components: 2 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/creative_writing
|
1,489 ±4.58 | 1,519 ±5 | 61.5% | 2026-10-02 | high |
| LMArena |
text/longer_query
|
1,506 ±3.48 | 1,499 ±3 | 32.2% | 2026-10-02 | 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 Language: 繁體中文 1,474 [1,464–1,484] Percentile 98 · #6 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/chinese
|
1,551 ±6.98 | 1,494 ±6 | 93.4% | 2026-10-02 | 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 Language: Deutsch 1,444 [1,422–1,466] Percentile 89 · #17 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/german
|
1,479 ±15 | 1,470 ±13 | 90.3% | 2026-10-02 | 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 Language: English 1,474 [1,467–1,480] Percentile 97 · #8 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/english
|
1,498 ±3.53 | 1,492 ±4 | 93.8% | 2026-10-02 | 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 Language: Español 1,455 [1,437–1,472] Percentile 95 · #8 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/spanish
|
1,485 ±10 | 1,477 ±10 | 91.9% | 2026-10-02 | 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 Language: Français 1,455 [1,437–1,474] Percentile 93 · #10 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/french
|
1,500 ±11 | 1,479 ±10 | 91.5% | 2026-10-02 | 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 Language: 日本語 1,477 [1,456–1,497] Percentile 96 · #6 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/japanese
|
1,503 ±15 | 1,504 ±12 | 91.0% | 2026-10-02 | 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 Language: 한국어 1,462 [1,443–1,482] Percentile 96 · #6 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/korean
|
1,468 ±14 | 1,487 ±11 | 91.2% | 2026-10-02 | 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 Language: Polski 1,467 [1,444–1,490] Percentile 96 · #6 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/polish
|
1,505 ±15 | 1,496 ±13 | 90.1% | 2026-10-02 | 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 Language: Русский 1,486 [1,477–1,496] Percentile 99 · #4 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/russian
|
1,510 ±5.48 | 1,506 ±5 | 93.5% | 2026-10-02 | 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
History
The same in a table
| Model | 2026-10 |
|---|---|
| Gemini 3.8 Flash | 1,472 S |
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.