MiniMax H3 Image local image editing: 1-min edits but 25-min prompt encoding
Front_Praline9683 · reddit · 2026-11-11
A Reddit user reports a split experience running MiniMax H3 Image locally for image editing: the 4-step edit on a 16MP image takes only 1 minute, but with just one reference image the prompt encoding takes 25 minutes to generate 8k tokens — a major bottleneck.
Setup: default template qwen3.5 9b w4a8 + abliterate LoRA. They're asking for faster alternative models, or a way to plug in an OpenAI-compatible API (gemini, glm, gpt, nemotron, qwen, etc.) and uncensor the model.
More from Infra
- Polymarket puts 17% odds on orbital AI data centers by 2027 as Google launches Suncatcher — Polymarket · 2026-11-11
- Developer drops $10B Michigan data center after residents reject $54M tax promise — Polymarket · 2026-11-11
- Rumor: DeepSeek to pretrain ~2T-param V5 on Huawei Ascend 950, release may slip to 2027 — teortaxesTex · 2026-11-11
- WSJ: Masayoshi Son Over-Reserved a Big Chunk of the Gas Turbine Market in AI Frenzy — mattwbaker · 2026-11-11
- FT: Commerce Dept stalls Megaspeed export blacklist despite 4 agencies pushing over Nvidia chip smuggling — rohanpaul_ai · 2026-11-11
- Run Pi Coding Agent Locally with TensorSharp + Qwen3.8-27B: Full CLI & SDK Tutorial — fuzhongkai · 2026-11-11