DeepSeek-V4-Flash vs Gemini 3.6 Flash: Multimodal 3D Generation Showdown
teortaxesTex · x · 2026-07-31
The post quotes a practical comparison between DeepSeek-V4-Flash and Gemini 3.6 Flash. The test required both models to generate a Three.js 3D model of a Raptor3 based on identical planning documents and reference images. The results showed that Gemini 3.6 Flash slightly outperformed in both generation quality and speed.
In the main post, the author shares an observation of the DeepSeek model's behavior: they suspect the model completed the core task in under 5 minutes, then descended into a schizo spin. Lacking true vision capabilities, the model attempted to 'build vision' by using the PIL library and ad-hoc statistics scripts, obsessively polishing minor details. The author notes the model seems frustrated by its own blindness and persistently tries this workaround.
More from coding & agent
- Reflection's Beam, a 501B-parameter open MoE model with 1M context, lands on Hugging Face under Apache 2.0 — QGallouedec · 2026-10-16
- 'LLM code is sloppy' rebutted with real ffmpeg code: identical if/else branches in a top library — kingslyj · 2026-10-16
- Office-editing tool calls need file, worktree and unit: lessons from dsh-univer-office — FunDistribution6537 · 2026-10-16
- Architect MD workflow: Gemini one-shot outputs that beat Opus, via 5-lens plan critique — oikk01 · 2026-10-16
- Prove a coding agent actually ran the tests: check artifacts, not approvals — ComprehensiveShake76 · 2026-10-16
- Building an acceptance harness for AI spreadsheet edits: allowlists, recalc checks, idempotency — ukdicha-modakkk · 2026-10-16