Models/gpt-6-luna

GPT 6 Luna

14th on the current Next.js board at high with 81% (90% with AGENTS.md). 6th on Supabase without skills at 70%, 77% with skills. Seventh on the Android long board at 10.7%, run cost $5.3.

OpenAI · Proprietary · Fetched 2026-10-10

LMArena WebDev

1579

#41 / 153 · ±6

LMArena Image-to-WebDev

1571

#25 / 65 · ±10

Android Bench

10.7%

#9 / 11 · 4.0–17.3

Benchmark scores

Boards older than 90 days are not counted as recommendations

BenchmarkSettingScoreRankCostAs of

LMArena WebDevmax1579±6#41—2026-10-27

LMArena Image-to-WebDevmax1571±10#25—2026-10-19

Android Benchcodex10.74.0–17.3#9$5.32026-10-26

Terminal-Bench 4.0max · Codex16.4±2.7#31$72.142026-10-06

The Slop Index—58#31—2026-10-28

Supabase Evalsmedium · Codex70.0#6—2026-10-30

Next.js Agent Evalshigh · Codex81.0#14$0.0122026-10-30

Convex Evals4 runs67.19#24$0.084762026-10-28

FrontierCodemax · codex21.9#16$0.1742026-10-27

SWE-bench Verified—This board has stopped updating——2026-02-26

Aider Polyglot—This board has stopped updating——2025-11-20

Use it or not

Why

Next.js baseline moved from 77% to 81% at about $0.012.
A $5.3 long-horizon run, the lowest cost on the Android board.

When not

Android pass rate 10.7%; still last on the public Supabase no-skills board. Do not leave multi-module work to it alone.

Appears in these scenarios

Collected from scenario boards

Build a Web frontend#14Next.js 81.0%Build an Android app#9Android Bench 10.7%Write CLI, scripts, and DevOps#16FrontierCode 21.9%Write docs & copy#31Slop Index 58

Peers

ModelWebDevPrice $/M

GPT 6 LunaCurrent1579—DeepSeek V4 Pro1582—DeepSeek V4 Flash1582—