Qwen3.8 27B
Qwen ·
Large dense multimodal model with frontier-class agentic coding capability for its size.
- 27B dense
- 256K context
- Vision
- DFlash2
- Apache 2.0
59% Intelligence
- Q420.5 GB
- Q628.3 GB
- Q834.4 GB
Browse open models
Every model in the Magnitude catalog, tuned and tested for Apple Silicon, NVIDIA, AMD, and CPU.
15 models
Gemma 4 31B
Google ·
Large dense Gemma model with the strongest measured coding score in its family.
- 31B dense
- 256K context
- Vision
- Apache 2.0
33% Intelligence
- Q4 QAT18.5 GB
Qwen3.6 35B-A3B
Qwen ·
Efficient MoE coding model with a large knowledge footprint and low active compute.
- 35B · 3B active
- 256K context
- Vision
- DFlash
- Apache 2.0
32% Intelligence
- Q424.2 GB
- Q528.5 GB
- Q633.9 GB
- Q840.4 GB
Muse Glimmer 30B
Meta ·
Dense multimodal model purpose-built for autonomous agentic tasks on consumer hardware.
- 30B dense
- 128K context
- Vision
- DFlash
- Apache 2.0
30% Intelligence
- Q419.6 GB
- Q525.5 GB
- Q629.9 GB
- Q836.0 GB
Gemma 4 26B-A4B
Google ·
Mid-size MoE model balancing a substantial weight footprint with low active compute.
- 25B · 3.8B active
- 256K context
- Vision
- Apache 2.0
29% Intelligence
- Q4 QAT15.4 GB
Gemma 4 12B
Google ·
Mid-size dense model with native tool use, reasoning, and multimodal capability.
- 12B dense
- 256K context
- Vision
- Apache 2.0
25% Intelligence
- Q4 QAT6.9 GB
Qwen3.5 4B
Qwen ·
Compact dense model for machines where responsiveness and footprint matter most.
- 4B dense
- 256K context
- Vision
- Apache 2.0
23% Intelligence
- Q43.7 GB
- Q54.0 GB
- Q64.9 GB
- Q86.7 GB
MiniCPM5 2B
OpenBMB ·
Compact dense model with native long context, tool use, and reasoning for on-device work.
- 2.5B dense
- 128K context
- DSpark
- Apache 2.0
22% Intelligence
- Q42.2 GB
- Q83.3 GB
Nemotron 3.5 Lightning 30B-A3B
NVIDIA ·
Efficient hybrid MoE model for local reasoning, coding, and agentic workflows.
- 30B · 3B active
- 1M context
- DFlash
- OpenMDW 1.1
22% Intelligence
- NVFP4 QAT23.6 GB
- Q426.6 GB
- Q836.2 GB
Qwen3.5 9B
Qwen ·
Small dense model with a substantial capability gain over the 4B tier.
- 9B dense
- 256K context
- Vision
- Apache 2.0
19% Intelligence
- Q47.1 GB
- Q57.8 GB
- Q69.9 GB
- Q814.2 GB
Gemma 4 E4B
Google ·
Small dense model with per-layer embeddings that materially improves capability over the E2B tier while remaining practical on memory-constrained machines.
- 8B dense
- 128K context
- Vision
- Apache 2.0
15% Intelligence
- Q4 QAT5.2 GB
Liquid LFM2.5 2.6B
Liquid AI ·
Compact dense Liquid model tuned for fast on-device tool use and coding workflows.
- 2.6B dense
- 128K context
- DSpark
- LFM 1.0
15% Intelligence
- Q42.0 GB
- Q52.3 GB
- Q62.6 GB
- Q83.2 GB
MiniCPM5 1B
OpenBMB ·
Compact dense model for on-device reasoning and tool use.
- 1.1B dense
- 128K context
- Apache 2.0
15% Intelligence
- Q40.7 GB
- Q81.2 GB
Gemma 4 E2B
Google ·
Very small dense model with per-layer embeddings optimized for on-device use.
- 5.1B dense
- 128K context
- Vision
- Apache 2.0
14% Intelligence
- Q4 QAT3.6 GB
Liquid LFM2.5 8B-A1B
Liquid AI ·
Low-active-parameter Liquid MoE optimized for fast local reasoning and tool use.
- 8.3B · 1.5B active
- 125K context
- DSpark
- LFM 1.0
13% Intelligence
- Q45.5 GB
- Q56.4 GB
- Q67.3 GB
- Q89.4 GB
No models match. Clear a filter to see more.
How to read this catalog
What is the intelligence score?
How a model compares with the most capable model available. It is the model’s Artificial Analysis Intelligence Index score, a composite of reasoning, coding, math, and knowledge benchmarks, as a percentage of the top model on the index. Where Artificial Analysis has not fully measured a model, Magnitude uses an estimate.
What do Q4, Q5, Q6, and Q8 mean?
Quantization levels. Lower numbers mean smaller downloads and less memory at a small cost in fidelity. Variants marked QAT were trained with quantization in mind and keep more fidelity at the same size. Magnitude recommends the highest-fidelity variant that fits your memory with room for context.
How are download sizes measured?
Weights plus every companion file Magnitude installs: the vision projector for image-capable models and the draft model for speculative decoding. They are the exact files pinned by the catalog’s locked Hugging Face commits.
Will these run on my machine?
Magnitude profiles your hardware and estimates tokens per second for every model before you download. The sizes here are what you download; the app gives you the real answer for your chip and memory, including room for context.
Catalog data from magnitudedev/magnitude, updated Sep 29, 2026.