# GPT-6.2 Astra & Sol benchmarks: OpenAl nailed the speed/cost curve, tied #1 on Epoch ECI, and kinda regressed in coding. — retained record

> Retained reddit record from reddit:OpenAI.

Canonical: https://highsignal.app/data/records/6abd5c3505025651

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:OpenAI)  published 2026-11-17 20:16 ingested 2026-11-18 02:38

[original source ↗](https://www.reddit.com/r/OpenAI/comments/1xyuzly/gpt62_astra_sol_benchmarks_openal_nailed_the/)

## retained text



```
Dug through the Artificial Analysis and Epoch benchmark dump for GPT-6.2 Astra and GPT-6.2 Sol. and the efficiency numbers are wild, but one catch for coders 👀 

• **GPT-6.2 Astra** owns the "Most Attractive Quadrant" among frontier models: Compared to the other flagship models plotted  (Fable 5.2, Opus 5.5, and Gemini 4 Carbon), GPT-6.2 Astra (max) is the only one sitting inside the green "Most attractive quadrant" for both speed and cost. It scores 62 on the Intelligence Index at \~10.5 minutes per task. Meanwhile, Claude Fable 5.2 (max) edges it out on intelligence (65), but costs 6.6x more and takes \~41 minutes (\~4x slower). 

• **Tied for #1 on Epoch ECI**: GPT-6.2 Astra jumped 5 points over GPT-6 Astra (169 → 174) to tie Claude Fable 5.2 (174) at the top of the leaderboard.   
Bullet GPT-6.2 Sol is a budget sleeper: GPT-6.2 Sol (max) hits 57 intelligence—right behind Claude Opus 5.5 (59) and Gemini 4 Carbon (60)—while running in \~8 minutes. 

• **regressed in coding**: Coding Agents stalled and regressed: On the AA Coding Agent Index, Codex - GPT-6.2 Astra (max) stayed completely flat at 53, matching GPT-6 Astra and trailing Claude Fable 5.2 (58) and Gemini 4 Carbon (55). Meanwhile, Codex - GPT-6.2 Sol (max) actually dropped 3 points from GPT-6.1 Sol (52 → 49) 👀.
```



document fetched 2026-11-18 00:28 · canonical https://reddit.com/r/OpenAI/comments/1xyuzly/gpt62_astra_sol_benchmarks_openal_nailed_the
