RangerIO Local Model Matrix
⏳ Detecting…
VRAM and RAM are editable — type your actual values for accurate results. Or pick a GPU preset above.
45 models
?
Qwen 3 0.6BFast Tier
Alibaba·600M·Dense·32K ctx
Unknown
chatedgeApache 2.0 · 2025-04
?
Qwen 3.5 0.8BFast Tier
Alibaba·800M·Dense·32K ctx
Unknown
chatedgeApache 2.0 · 2026-02
?
Gemma 3 1B
Google·1B·Dense·32K ctx
Unknown
chatedgeGemma · 2025-03
?
DeepSeek R1 1.5B
DeepSeek·1.5B·Dense·64K ctx
Unknown
reasoningMIT · 2025-01
?
Qwen 3 1.7B
Alibaba·1.7B·Dense·32K ctx
Unknown
chatmultilingualApache 2.0 · 2025-04
?
Qwen 3.5 2B
Alibaba·2B·Dense·32K ctx
Unknown
chatmultilingualApache 2.0 · 2026-02
?
SmolLM3 3B
HuggingFace·3B·Dense·128K ctx
Unknown
chatreasoningApache 2.0 · 2025-07
?
IBM Granite 4.0 MicroDefault
IBM·3B (1B active)·MoE·128K ctx
Unknown
chatcomplianceedgeApache 2.0 · 2025-10
?
Phi-3.5 Mini
Microsoft·3.8B·Dense·128K ctx
Unknown
reasoningcodechatMIT · 2024-08
?
Phi-4 Mini Reasoning
Microsoft·3.8B·Dense·16K ctx
Unknown
reasoningMIT · 2025-04
?
Qwen 3 4B
Alibaba·4B·Dense·32K ctx
Unknown
chatcodeApache 2.0 · 2025-04
?
Gemma 3 4B
Google·4B·Dense·128K ctx
Unknown
chatvisionGemma · 2025-03
?
Qwen 3.5 4B
Alibaba·4B·Dense·32K ctx
Unknown
chatmultilingualApache 2.0 · 2026-02
?
IBM Granite 4.0 3B Vision
IBM·4B·Dense·32K ctx
Unknown
visioncomplianceApache 2.0 · 2026-03
?
Gemma 4 E2B
Google·5B·Dense·256K ctx
Unknown
chatvisionGemma · 2026-04
?
DeepSeek R1 Distill 7B
DeepSeek·7B·Dense·64K ctx
Unknown
reasoningMIT · 2025-01
?
IBM Granite 4.0 Tiny
IBM·7B (1B active)·MoE·128K ctx
Unknown
chatcompliancereasoningApache 2.0 · 2025-10
?
Qwen 3 8B
Alibaba·8B·Dense·128K ctx
Unknown
chatcodereasoningApache 2.0 · 2025-04
?
Qwen 3.5 9BRecommended
Alibaba·9B·Dense·32K ctx
Unknown
chatvisionApache 2.0 · 2026-02
?
Nemotron Nano 9B v2
NVIDIA·9B·Dense·128K ctx
Unknown
reasoningNVIDIA Open · 2025-06
?
Gemma 3 12B
Google·12B·Dense·128K ctx
Unknown
chatvisionreasoningGemma · 2025-03
?
Qwen 3 14B
Alibaba·14B·Dense·128K ctx
Unknown
chatcodereasoningApache 2.0 · 2025-04
?
DeepSeek R1 Distill 14B
DeepSeek·14B·Dense·64K ctx
Unknown
reasoningMIT · 2025-01
?
Phi-4 14B
Microsoft·14B·Dense·16K ctx
Unknown
reasoningcodeMIT · 2024-12
?
GPT-OSS 20B
OpenAI·21B (3.6B active)·MoE·128K ctx
Unknown
chatreasoningcodeApache 2.0 · 2025-08
?
LFM2 24B
Liquid AI·24B (2.3B active)·MoE·32K ctx
Unknown
chatedgeragLiquid AI · 2025-11
?
Gemma 3 27B
Google·27B·Dense·128K ctx
Unknown
chatvisionreasoningGemma · 2025-03
?
Gemma 4 26B-A4B
Google·27B (4B active)·MoE·256K ctx
Unknown
chatvisionreasoningGemma · 2026-04
?
Qwen 3.5 27B
Alibaba·27.8B·Dense·256K ctx
Unknown
chatvisionreasoningApache 2.0 · 2026-02
?
Qwen 3 30B-A3B
Alibaba·30B (3.3B active)·MoE·128K ctx
Unknown
chatreasoningApache 2.0 · 2025-04
?
Nemotron 3 Nano 30B
NVIDIA·30B (3B active)·MoE·1M ctx
Unknown
chatreasoningNVIDIA Open · 2025-06
?
Qwen 3 32B
Alibaba·32B·Dense·128K ctx
Unknown
chatcodereasoningApache 2.0 · 2025-04
?
DeepSeek R1 Distill 32B
DeepSeek·32B·Dense·64K ctx
Unknown
reasoningMIT · 2025-01
?
Gemma 4 31B
Google·33B·Dense·256K ctx
Unknown
chatvisionreasoningGemma · 2026-04
?
Qwen 3.5 35B-A3B
Alibaba·35B (3B active)·MoE·256K ctx
Unknown
chatvisionApache 2.0 · 2026-02
?
Llama 3.3 70B
Meta·70B·Dense·128K ctx
Unknown
chatreasoningcodeLlama 3.3 · 2024-12
?
Llama 4 Scout 17B
Meta·109B (17B active)·MoE·128K ctx
Unknown
chatvisionreasoningLlama 4 · 2025-04
?
GPT-OSS 120B
OpenAI·117B (5.1B active)·MoE·128K ctx
Unknown
chatreasoningcodeApache 2.0 · 2025-08
?
Qwen 3.5 122B-A10B
Alibaba·122B (10B active)·MoE·256K ctx
Unknown
chatvisionreasoningApache 2.0 · 2026-02
?
Qwen 3 235B-A22B
Alibaba·235B (22B active)·MoE·128K ctx
Unknown
chatcodereasoningApache 2.0 · 2025-04
?
Qwen 3.5 397B-A17B
Alibaba·397B (17B active)·MoE·256K ctx
Unknown
chatvisionreasoningcodeApache 2.0 · 2026-02
?
Llama 4 Maverick
Meta·400B (17B active)·MoE·1M ctx
Unknown
chatvisionreasoningcodeLlama 4 · 2025-04
?
DeepSeek R1
DeepSeek·671B (37B active)·MoE·64K ctx
Unknown
reasoningMIT · 2025-01
?
DeepSeek V3.2
DeepSeek·685B (37B active)·MoE·128K ctx
Unknown
chatcodereasoningMIT · 2025-12
?
Kimi K2
Moonshot AI·1T (32B active)·MoE·128K ctx
Unknown
chatreasoningcodeKimi · 2025-07
RangerIO — Privacy-first AI data intelligence. 100% on-device.
No cloud. No data sharing. No new compliance surface.
VRAM estimates based on GGUF quantized models via llama.cpp. Actual requirements may vary.
No cloud. No data sharing. No new compliance surface.
VRAM estimates based on GGUF quantized models via llama.cpp. Actual requirements may vary.