3D beach diorama test: Opus 5.5 first, DeepSeek V4.1 Flash burns 124M tokens for $2.47
In a shared 3D beach diorama build, Opus 5.5 ranked first at ~1h50m and ~$29, GPT-6.1 Sol second at ~2h51m and ~$5.35, while DeepSeek V4.1 Flash 1017 ran 5 parallel agents for 4h19m and 124M tokens at just $2.47, with the author saying a few more fix prompts could close the gap.
Ran a test pitting Claude Opus 5.5 against GPT-6.1 Sol and the new DeepSeek V4.1 Flash (1017) on a complex 3D Beach Diorama build (including custom water shader reworks, animations & wildlife) 🏖️🌊⚡️ Actual session audit stats (factoring in all fixes): 🥇 1. Claude Opus 5.5 • Active working time: ~1h 50m (209 model calls, 249 tool calls) • Tokens: 49.4M total (47.4M cache read, 1.6M cache write, 338k output) • API cost equiv.: ~$29 (1-hour cache expiry makes follow-up edits after breaks pretty pricey) 🥈 2. GPT-6.1 Sol • Active working time: 2h 51m • Tokens: 16.43M total (15.65M cached, 596k uncached, 185k output) • API cost equiv.: ~$5.35 (Standard tier) 🥉 3. DeepSeek V4.1 Flash (1017 — 5 parallel agents via OpenCode Go) • Active working time: 4h 19m (~8h total LLM time across 5 agents, 935 calls!) • Tokens: 124.42M total (114.6M cached, 1.45M My take: Opus 5.5 grabs #1 for quality and speed, right ahead of Sol 6.1. That being said, DeepSeek 4.1 Flash is genuinely impressive — churning through 124M tokens for $2.47 is insane, and with just a couple more fix prompts, it could easily close the gap with the top two 🔥 Try all three live right in your browser 👇 🔗 Opus 5.5: https://beach-3d-opus.vercel.app/ 🔗 Sol 6.1: https://beach-3d-sol.vercel.app/ 🔗 DeepSeek 4.1 Flash: https://beach-3d-deepseek.vercel.app/