MiniMax-H3-Image-PE
- Open weights
Prices
0 offerings · prices as listedNo priced offering is listed for this model yet.
Details
- Reasoning efforts
- xhigh, medium, low
- Licence
- minimax-h3-image-community-license-agreement
- Family
- minimax-h-image-pe
Not stated on MiniMax’s pages: max input, input modalities, output modalities, reasoning, max thinking budget, tool use, structured output, built-in tools, knowledge cutoff, aliases.
Run locally
Weights: MiniMaxAI/MiniMax-H3-Image-PE- Parameters 27.4B
- Dense
- Native precision BF16
- Layers 64
- KV cache per token 64 KB
- Config positions 256K (config.json max_position_embeddings)
- Needs
- ≈ 18.9 GB (weights 17.5 + KV cache 0.5 + overhead 0.9)
- Usable
- 21.6 GB (90% of its memory)
- Generation speed
- 26–39 tok/s
- Longest context
- 51K (longer doesn’t fit)
- First token
- about 0.88–1.32 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native) | 51.0 GB | ≈ 54.0 | ✕Doesn’t fit | — |
| Q8_0 | 27.1 GB | ≈ 28.9 | ○Needs system memory 4–7 tok/s | — |
| FP8 | 25.5 GB | ≈ 27.3 | ○Needs system memory 5–8 tok/s | — |
| Q6_K | 20.9 GB | ≈ 22.5 | ○Needs system memory 15–23 tok/s | — |
| Q5_K_M | 18.1 GB | ≈ 19.5 | ●Runs well 25–38 tok/s | 41K |
| Q5_0 | 17.5 GB | ≈ 18.9 | ●Runs well 26–39 tok/s | 51K |
| Q4_K_M | 15.5 GB | ≈ 16.8 | ●Runs well 29–44 tok/s | 85K |
| Q4_0 | 14.3 GB | ≈ 15.5 | ●Runs well 32–47 tok/s | 105K |
| IQ4_XS | 13.5 GB | ≈ 14.7 | ●Runs well 33–50 tok/s | 118K |
| Q3_K_Mquality drops | 12.5 GB | ≈ 13.6 | ●Runs well 36–54 tok/s | 136K |
| IQ3_XXSquality drops | 9.7 GB | ≈ 10.7 | ●Runs well 46–69 tok/s | 181K |
| Q2_Kquality drops | 9.6 GB | ≈ 10.6 | ●Runs well 47–70 tok/s | 184K |
| IQ2_XXSquality drops | 6.6 GB | ≈ 7.6 | ●Runs well 66–100 tok/s | 233K |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU GeForce RTX 4090; unified memory Apple M6 (24GB or 32GB) · 32 GB.
Run commands
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-11. How it’s estimated
Cited scores
From third-party leaderboards, not measured by usThis model isn’t on any leaderboard we cite yet.
Changes
All changes →- New language model
Sources
- MiniMaxAI/MiniMax-H3-Image-PE model card (opens in a new tab) huggingface.co · checked Nov 10, 2026
- MiniMaxAI/MiniMax-H3-Image-PE config.json (opens in a new tab) huggingface.co · checked Nov 10, 2026
- MiniMaxAI/MiniMax-H3-Image-PE model card (opens in a new tab) huggingface.co · checked Nov 10, 2026
MiniMax’s official policies: data use, privacy, terms →
Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data
MiniMax · China · minimax.io
Related models
Snapshots, the same family and other models from this provider