Blender agent test: Opus 5.5 beats GPT-6 Astra in 73 min at $27.3 vs 88 min at $21.3
rms80 · x · 2026-09-23
A same-prompt agent benchmark: models drove Blender via the Higgsfield plugin and MCP (with Nano Banana Pro for images, Tripo for 3D) to build the DeLorean and clock tower square, rig the car, and animate the lightning strike.
- Opus 5.5: 73 min, 218k output tokens, $27.3 API cost. Working doors, spinning wheels, strike lands.
- GPT-6 Astra: 88 min, 136k output tokens, $21.3. Skipped generators and hand-built all 517 objects.
The poster notes how quickly the "Astra-pilled" and "Fable" hype cycles have faded.
More from Models
- Cheap per token, expensive per task: charting AI model pricing vs. task performance — Wsz2020 · 2026-09-23
- 6 luna models put to the drawing test via computer use — results not bad — adonis_singh · 2026-09-23
- Computer use drawing test: Opus vs Astra recreating a reference image — adonis_singh · 2026-09-23
- GPT-Live-1 wins at Mafia by persuading humans to vote out rival players — pbbakkum · 2026-09-23
- Early Hands-On: Opus 5.5 Called 'Sooo Good' to Talk To in First Impressions — daniel_mac8 · 2026-09-23
- Model profitability analysis: Opus 5.5 beats Fable 5.1 at half the price, Grok loses on every task — Wsz2020 · 2026-09-23