Opus 5 adds numeric self-checks to the Boeing benchmark and outbuilds Fable
victormustar · x · 2026-07-25
- The post argues that a comparison video makes it easier to see that Opus 5 produced a more detailed result than Fable on the Boeing benchmark.
- The quoted benchmark notes that Fable 5 took 42 minutes and produced about 4 files / 1,000 lines, using a Puppeteer harness, 12 camera presets, 17 render passes, and 54 screenshot inspections.
- Opus 5 took 71 minutes, but expanded the work to 20 modules / 4,000 lines and used a second tool, measure.mjs, to extract quantitative geometry from the running scene.
- That tool measured 14 aircraft metrics — including span, wheelbase, wing root leading-edge station, sweep, dihedral, and gear track — and compared them against published 747-400 figures with per-metric tolerances.
- The image and commentary together suggest Opus 5 does not just visually verify the scene; it tries to measure and self-check the generated output numerically.
Related event: Opus 5 Shows Superior Detail in Boeing Render Comparison(2 posts)→
More from Models
- FrontierML benchmark shows open models still far behind closed ones on end-to-end ML work — scaling01 · 2026-10-27
- Mirendil debuts Fern, its first from-scratch model with up to 10M-token context — shayan_ · 2026-10-27
- Week 3 of "Fable 5.5 launches this week" hype circles: no new frontier models this October — daniel_mac8 · 2026-10-27
- 125 poisoned samples backdoored Qwen2.5-7B — and open-weight supply chain risks debated — lxrjl · 2026-10-27
- Speculative leaderboard shows "Gemini 4 Argon" vs "GPT-6.1 Sol"; observer calls MiniMax parity suspicious — teortaxesTex · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27