jarrelscy/MiMo-V2.6-Pro-EXL3
text-generation model by jarrelscy
Older snapshot: check the original source for current information.
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
Experimental sequential quantization checkpoint. All 69 MoE layers are complete (layer 0 is dense), and it can be served with the custom loader and kernels at https://github.com/jarrelscy/mimo-v2.6-pro-exl3-sm120. It is not an off-the-shelf EXL3 model directory, so exllamav3/TabbyAPI, vLLM and SGLang cannot load it directly. Tested on 4x RTX PRO 6000 Blackwell (SM120, 96 GB, PCIe P2P, no NVLink): TP4, 72 GiB per GPU, 11.8 ms/step, 83 tok/s single-stream decode. Needs the CUDA 12.8 toolkit (the extensions are JIT-built on first start), Python 3.12 and 283 GB of disk. The server provides /v1/completions and /v1/chat/completions (streaming, reasoningcontent, tool calls). It serves one request at a time with a 32K context by default (MIMOLMAX). Text only: vision and audio are not wired up. MTP speculative decoding (MIMOSPEC=K) is experimental and off by default. See the GitHub README for the options. Completed full-corpus joint-PV layers: [1, 2, 3, 4, 5, 6, 7, 8,…
Read the full model card ↗ · Preview checked 2026-11-05T19:16:40.130Z
Inside the original model card — Document outline
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Documentation belongs to its respective authors. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
jarrelscy/MiMo-V2.6-Pro-EXL3 is a text-generation model repository published by jarrelscy on Hugging Face. The source reports the custom library.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- text-generation
- Library
- custom
- Recent downloads (30 days)
- 0
- Cumulative likes
- 0
- Hugging Face trending score
- 0
- Reported safetensors parameters
- Not reported
- Architecture
- Not reported
- License
- Not reported — check the model card
- Access
- Not gated by Hugging Face
- Created
- 2026-09-24T06:05:21.000Z
- Last modified
- 2026-09-24T06:23:56.000Z
Compatibility and lineage
Language coverage was not reported.
Reported base models: XiaomiMiMo/MiMo-V2.6-Pro-RL
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-09-24T06:37:32.927Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.