README.md

metadata

language:
  - en
  - zh
  - fr
  - es
  - pt
  - de
  - it
  - ru
  - ja
  - ko
  - ar
  - vi
  - th
license: apache-2.0
library_name: mlx
base_model: Qwen/Qwen3.6-35B-A3B
tags:
  - 4-bit
  - 4bit
  - apple-silicon
  - chat
  - conversational
  - edge-ai
  - function-calling
  - image-text-to-text
  - image-to-text
  - instruct
  - local-llm
  - m1
  - m2
  - m3
  - m4
  - mac
  - mac-mini
  - mac-studio
  - macbook-air
  - macbook-pro
  - macos
  - metal
  - mixture-of-experts
  - mlx
  - mlx-lm
  - mlx_vlm
  - mmlu-verified
  - moe
  - multimodal
  - no-cloud
  - offline
  - on-device
  - outlier
  - outlier-app
  - private
  - private-ai
  - quantized
  - qwen
  - qwen3.6
  - qwen3_5_moe
  - reasoning
  - safetensors
  - text-generation
  - thinking
  - tool-use
  - vision
  - visual-question-answering
  - vqa
pipeline_tag: image-text-to-text
model-index:
  - name: Outlier-Ai/Outlier-Vision-35B-A3B-MLX-4bit
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU (5-shot, n=14042)
          type: cais/mmlu
          config: all
          split: test
        metrics:
          - type: acc
            name: accuracy
            value: 0.8352
            verified: false
      - task:
          type: image-text-to-text
          name: Visual Question Answering
        dataset:
          name: HumanEval
          type: openai_humaneval
          split: test
        metrics:
          - type: pass@1
            name: pass@1
            value: 0.6098
            verified: false

Part of the Outlier shipping lineup. Outlier is a free macOS app that runs this model locally, with one click. Apple Silicon only.

Outlier Vision 35B-A3B (MLX 4-bit)

Multimodal MoE tier with image+text input (35B params, ~3.6B active per token). Optimized for image+text analysis, not code generation — use Core or Code for coding workflows.

Try it in Outlier

The simplest way to use this model is through the Outlier app — open the tier picker, select Outlier Vision, click download, and chat. No setup, no Python, no MLX install, no token quotas.

➡ Download Outlier — outlier.host

A screenshot of the tier picker is at outlier.host/screenshots/tier-picker.png.

Load this directly (power users)

If you want the raw MLX-4bit weights without the app:

pip install mlx-lm
python -m mlx_lm.generate \
  --model Outlier-Ai/Outlier-Vision-35B-A3B-MLX-4bit \
  --prompt "Write a quicksort in Python." \
  --max-tokens 512

from mlx_lm import load, generate
model, tokenizer = load("Outlier-Ai/Outlier-Vision-35B-A3B-MLX-4bit")
print(generate(model, tokenizer, prompt="Hello", max_tokens=256))

Verified benchmarks

For σ-qualified MMLU, HumanEval, and Mac inference-speed numbers — with full provenance (source file, command, n, stderr, date) — see outlier.host/benchmarks.

Other Outlier shipping tiers

• Outlier Nano 4B (entry tier, ~3 GB)
• Outlier Lite 9B (balanced, ~6 GB)
• Outlier Quick 26B-A4B MoE (~16 GB)
• Outlier Core 27B (default, ~16 GB)
• Outlier Code 27B (code-tuned, ~16 GB)

License

Apache 2.0 (inherits from upstream base model). Conversion artifact only — the underlying weights are governed by the base model's license.