GPT-6 Astra

By OpenAI. 11 results on 6 boards, copied as published. Different boards, different metrics: don't add them up.

License (as listed)Proprietary

Hosting regionNot confirmed

Boards6

Code

57.88%Terminal-Bench 4.0
GPT-6 Astra (xhigh) · Accuracy %
Codex · as of 2026-09-03 ↗
57.88%Terminal-Bench 4.0
GPT-6 Astra (high) · Accuracy %
Codex · as of 2026-09-03 ↗
54.24%Terminal-Bench 4.0
GPT-6 Astra (medium) · Accuracy %
Codex · as of 2026-09-03 ↗
50.61%Terminal-Bench 4.0
GPT-6 Astra (low) · Accuracy %
Codex · as of 2026-09-03 ↗

Agentic / terminal

58.18%Terminal-Bench 4.0
GPT-6 Astra (max) · Resolution rate
Codex · as of 2026-09-03 ↗

Websites

1774LMArena WebDev Arena
gpt-6-astra-max · Arena score · board #2
LMArena WebDev (agentic front-end tasks) · as of 2026-10-22 ↗
89.59%Vals AI Vibe Code Bench v1.1
GPT-6 Astra · Score % · board #7
OpenHands · as of 2026-10-20 ↗
1357Design Arena (Fullstack)
gpt-6-astra · Elo · board #1
Design Arena · as of 2026-10-24 ↗

Design

1383Design Arena (All categories)
gpt-6-astra · Elo · board #2
Design Arena · as of 2026-10-24 ↗
1371Design Arena (All categories)
gpt-6-astra-max · Elo · board #4
Design Arena · as of 2026-10-24 ↗

Android

28%Android Bench 2.0 (long-horizon, 30 tasks)
GPT 6 Astra · Pass rate
codex · as of 2026-10-06 ↗

iOS

No mature public leaderboard

Available in

ChatGPT Plus ChatGPT Pro

Plans whose row in our pricing data names this model.

Use cases with this model

GPT-6 AstraClaude Opus 5.5Fable 5.1+1Model test

Test: "Muse Wonder Full 1.0 on real work. 2 repos. 105 bugs … GPT-6 Astra (max): 45 Muse Wonder Full 1.0 (max): 44 Fable 5.1 (max): 43 Opus 5.5…

Builder: @PawelHuryn
Posted: Oct 13, 2026
Plan: —
Cost: Not disclosed

CodeRead the post

x.com · link re-checked 2026-10-16

GPT-6 AstraClaude Opus 5.5

"Lit-Moss-Building": GPT-6 Astra designed it, Claude Opus 5.5 built it in three.js

Builder: @chetanankola
Posted: Oct 10, 2026
Plan: —
Cost: Not disclosed

DesignRead the post

x.com · link re-checked 2026-10-16

GPT-6 AstraClaude Opus 5.5Kimi K3

MotionPromptGallery: "178 AI motion graphics … made with Opus 5.5, GPT-6 Astra, Kimi K3, Fable 5", Next.js on Vercel

Builder: @mikenevermiss
Posted: Oct 9, 2026
Plan: —
Cost: Not disclosed

WebsitesDesignRead the post

x.com · link re-checked 2026-10-16

GPT-6 AstraClaude Opus 5.5

A one-shot interactive visualization of Calvino's Invisible Cities

Builder: Piotr Migdał
Posted: Oct 7, 2026
Plan: API
Cost: Stated: Opus 5.5: about $74 in API tokens, 1h25m. GPT-6 Astra: about $10, 53 min at medium effort

DesignRead the post

quesma.com · web source · link re-checked 2026-10-16

DeepSeek V4 ProGLM 5.3GPT-6 Astra+5

Same 3D prompt one-shot across 8 models incl. Mistral Large 4

Builder: @domsilvestre
Posted: Oct 6, 2026
Plan: —
Cost: Not disclosed

DesignRead the post

x.com · link re-checked 2026-10-16

GPT-6 AstraGemini 4 ArgonModel test

Test: an isometric dashboard compared across GPT-6 Astra and Gemini 4 Argon. "Argon packs in more detail"

Builder: @AIBoticssq
Posted: Oct 6, 2026
Plan: —
Cost: Not disclosed

DesignRead the post

x.com · link re-checked 2026-10-16

GPT-6 AstraClaude Opus 5.5

StarSkirmish Hillclimb: LLMs write C++ StarCraft: BW bots (BWAPI) with no time limit

Builder: __cayenne__
Posted: Oct 2, 2026
Plan: GPT-6 Astra, Claude 5.5 Opus
Cost: Not disclosed

CodeRead the post

news.ycombinator.com · web source · link re-checked 2026-10-16

GPT-6 Astra

A from-scratch "recreation of Zelda: Ocarina of Time … After 500+ hours of agent runtime … Built with GPT 6 Astra"

Builder: @LexnLin
Posted: Sep 25, 2026
Plan: —
Cost: Not disclosed

CodeDesignRead the post

x.com · link re-checked 2026-10-16

GPT-6 AstraClaude Opus 5.5Kimi K3Model test

Cost (test): "Opus 5.5: $8.95 | 10 mins GPT-6 Astra: $7.45 | 8 mins GPT-6 Sol: $3.90 | 5 mins Kimi K3: $4.04 | 6 mins" (Eiffel Tower in three.js)

Builder: @Bhavani_00007
Posted: Sep 23, 2026
Plan: —
Cost: Not disclosed

DesignRead the post

x.com · link re-checked 2026-10-16

OpenAI CodexGPT-6 AstraGPT-5.6 Sol

Created this in just 10 min using GPT-6 Astra … built entirely in Codex - 90% GPT-6 Astra, rest GPT-5.6 Sol

Builder: @himanshubuildss
Posted: Sep 22, 2026
Plan: —
Cost: Not disclosed

WebsitesDesignRead the post

x.com · link re-checked 2026-10-16

ChatGPTGPT 6 Astra

"I built this entire app with ChatGPT 6 Astra in just three hours"

Builder: @UniverseIce
Posted: Sep 11, 2026
Plan: —
Cost: Not disclosed

CodeRead the post

x.com · link re-checked 2026-10-16

GPT-6 AstraKimi K3

Chess game pitting GPT-6 Astra vs Kimi K3 via API ("Kimi K3 … beat it")

Builder: @Bhavani_00007
Posted: Sep 9, 2026
Plan: —
Cost: Not disclosed

CodeRead the post

x.com · link re-checked 2026-10-16

Same family: GPT 13

GPT-6.1 SolClaude Opus 5.5

A game "in about two weeks". GPT-6.1 Sol made the assets, Opus 5.5 did "most of the coding and game logic"

Builder: @hakimieiqbal
Posted: Oct 15, 2026
Plan: —
Cost: Not disclosed

CodeRead the post

x.com · link re-checked 2026-10-16

GPT-6.1 SolMuse WonderModel test

Test: "Muse Wonder Full Max vs GPT-6.1 Sol Ultra. For me, Muse Wonder is the clear winner" on physics realism

Builder: @SPAC89
Posted: Oct 12, 2026
Plan: —
Cost: Not disclosed

DesignRead the post

x.com · link re-checked 2026-10-16

GPT family

GPT 6.1 Sol design … One shot, without design review. It's not bad tbh.

Builder: @Dimillian
Posted: Oct 8, 2026
Plan: —
Cost: Not disclosed

DesignRead the post

x.com · link re-checked 2026-10-16

GPT 6.1 SolMistral large 4

Cost (user-reported): "Mistral large 4 costs $14 vs GPT 6.1 $1" for the same revenue-forecast coding job

Builder: @victor_zhng
Posted: Oct 7, 2026
Plan: —
Cost: Not disclosed

CodeRead the post

x.com · link re-checked 2026-10-16

OpenAI CodexGPT family

A Snew landing page update: "For the first time in my life, I built something directly using Codex only."

Builder: @joaopaulots
Posted: Oct 7, 2026
Plan: —
Cost: Not disclosed

WebsitesRead the post

x.com · link re-checked 2026-10-16

Claude CodeGPT family

I recorded a 45-min tutorial on how to create paid apps with Claude Code and Codex … I hit $100K MRR

Builder: @MengTo
Posted: Oct 7, 2026
Plan: —
Cost: Not disclosed

WebsitesRead the post

x.com · link re-checked 2026-10-16

All 13 GPT use cases →

Sources

lmarena.ai/leaderboard/webdev vals.ai/benchmarks/vibe-code developer.android.com/bench tbench.ai/leaderboard designarena.ai/leaderboard designarena.ai/leaderboard/fullstack

GPT-6 Astra

Available in

Use cases with this model

SLOP CITY: open-source first-person game that embeds the author's TikTok posts

Test: "Muse Wonder Full 1.0 on real work. 2 repos. 105 bugs … GPT-6 Astra (max): 45 Muse Wonder Full 1.0 (max): 44 Fable 5.1 (max): 43 Opus 5.5…

"Lit-Moss-Building": GPT-6 Astra designed it, Claude Opus 5.5 built it in three.js

Asana/StackAI CTO Frank Hidalgo had GPT-6 Astra in Codex refactor and run a 144-run study on the browser agent. Result: runs 76x cheaper and 5x faster. About 1 week instead of 1–2 months

MotionPromptGallery: "178 AI motion graphics … made with Opus 5.5, GPT-6 Astra, Kimi K3, Fable 5", Next.js on Vercel

A one-shot interactive visualization of Calvino's *Invisible Cities*

Same 3D prompt one-shot across 8 models incl. Mistral Large 4

Test: an isometric dashboard compared across GPT-6 Astra and Gemini 4 Argon. "Argon packs in more detail"

StarSkirmish Hillclimb: LLMs write C++ StarCraft: BW bots (BWAPI) with no time limit

Token-compression proxy for Codex (fine-tuned Qwen) that cuts tokens about 30%

A from-scratch "recreation of Zelda: Ocarina of Time … After 500+ hours of agent runtime … Built with GPT 6 Astra"

Cost (test): "Opus 5.5: $8.95 | 10 mins GPT-6 Astra: $7.45 | 8 mins GPT-6 Sol: $3.90 | 5 mins Kimi K3: $4.04 | 6 mins" (Eiffel Tower in three.js)

Created this in just 10 min using GPT-6 Astra … built entirely in Codex - 90% GPT-6 Astra, rest GPT-5.6 Sol

"I built this entire app with ChatGPT 6 Astra in just three hours"

Chess game pitting GPT-6 Astra vs Kimi K3 via API ("Kimi K3 … beat it")

Same family: GPT 13

A game "in about two weeks". GPT-6.1 Sol made the assets, Opus 5.5 did "most of the coding and game logic"

Test: "Muse Wonder Full Max vs GPT-6.1 Sol Ultra. For me, Muse Wonder is the clear winner" on physics realism

GPT 6.1 Sol design … One shot, without design review. It's not bad tbh.

Cost (user-reported): "Mistral large 4 costs $14 vs GPT 6.1 $1" for the same revenue-forecast coding job

A Snew landing page update: "For the first time in my life, I built something directly using Codex only."

I recorded a 45-min tutorial on how to create paid apps with Claude Code and Codex … I hit $100K MRR

Sources

The weekly diff

A one-shot interactive visualization of Calvino's Invisible Cities