GPT-6 Astra in fedi-index
The score in every slice, what it is made of, value, local run and history.
fedi-index · General
A 1,463 ±8
Percentile 95 · #16
People prefer / Solves tasks
1,437 / 1,514
Human votes vs benchmarks, Elo-eq
Value
+12
$20 per 1M tokens, median of 24 providers · vendor $20
On your hardware
Closed weights: cloud only.
Score decomposition
A General 1,463 [1,455–1,471] Percentile 95 · #16 · components: 9
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| Epoch AI |
eci
|
166 ±2.07 | 1,568 ±16 | 23.7% | 2026-09-03 | — |
| LMArena |
text/overall
|
1,442 ±3.49 | 1,442 ±4 | 18.7% | 2026-10-02 | max |
| LMArena |
text/hard_prompts
|
1,463 ±4.23 | 1,452 ±4 | 18.5% | 2026-10-02 | max |
| LMArena |
text/instruction_following
|
1,455 ±5.69 | 1,465 ±5 | 8.7% | 2026-10-02 | max |
| LMArena |
text/expert
|
1,492 ±10 | 1,470 ±8 | 7.6% | 2026-10-02 | max |
| Epoch AI |
simpleqa_verified
decay ×0.61 |
0.76 ±0.01 | 1,535 ±4 | 4.9% | 2026-08-30 | max |
| LMArena |
text/longer_query
|
1,458 ±5.12 | 1,455 ±5 | 4.5% | 2026-10-02 | max |
| LMArena |
text/multi_turn
|
1,452 ±8.60 | 1,446 ±8 | 3.9% | 2026-10-02 | max |
| Epoch AI |
gpqa_diamond
outlier decay ×0.26 |
0.96 ±0.01 | 1,485 ±6 | 1.0% | 2026-08-30 | max |
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 1,471 [1,467–1,474] Percentile 98 · #5 · components: 2 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
webdev/overall
|
1,788 ±5.65 | 1,532 ±2 | 60.7% | 2026-10-06 | max |
| LMArena |
text/coding
outlier |
1,489 ±6.62 | 1,460 ±6 | 26.6% | 2026-10-02 | max |
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 Agents 1,481 [1,472–1,491] Percentile 96 · #3 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
agent/overall
|
0.12 ±0.01 | 1,514 ±5 | 89.5% | 2026-10-02 | max |
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 1,478 [1,468–1,488] Percentile 98 · #6 · components: 3
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/math
|
1,469 ±14 | 1,459 ±13 | 37.5% | 2026-10-02 | max |
| Epoch AI |
frontiermath_t1_3
decay ×0.86 |
0.94 ±0.01 | 1,524 ±3 | 36.4% | 2026-08-30 | max |
| Epoch AI |
frontiermath_t4
decay ×0.86 |
0.98 ±0.02 | 1,548 ±4 | 17.8% | 2026-08-30 | 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 Vision 1,441 [1,434–1,448] Percentile 75 · #29 · components: 3 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
vision/overall
|
1,280 ±5.74 | 1,456 ±5 | 65.1% | 2026-10-02 | max |
| LMArena |
vision/ocr
|
1,292 ±6.70 | 1,464 ±6 | 15.4% | 2026-10-02 | max |
| LMArena |
vision/diagram
|
1,299 ±11 | 1,463 ±10 | 12.7% | 2026-10-02 | max |
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 1,434 [1,424–1,443] Percentile 80 · #54 · components: 2 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/creative_writing
|
1,419 ±7.11 | 1,448 ±7 | 59.7% | 2026-10-02 | max |
| LMArena |
text/longer_query
|
1,458 ±5.12 | 1,455 ±5 | 33.4% | 2026-10-02 | max |
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 1,424 [1,408–1,440] Percentile 75 · #56 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/chinese
|
1,487 ±11 | 1,443 ±9 | 92.2% | 2026-10-02 | max |
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 1,424 [1,414–1,434] Percentile 76 · #66 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/english
|
1,448 ±5.10 | 1,439 ±5 | 93.4% | 2026-10-02 | max |
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 1,425 [1,410–1,439] Percentile 77 · #55 · components: 1 provisional
| Source | Component | Value | Elo-eq | Weight | Measured | Variant |
|---|---|---|---|---|---|---|
| LMArena |
text/russian
|
1,440 ±8.83 | 1,443 ±8 | 92.6% | 2026-10-02 | max |
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 |
|---|---|
| GPT-6 Astra | 1,463 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.