Best local LLMs for the Orange Pi AIpro (20T) (24 GB)

Each list assumes the context that task needs, so a model that fits for chat may rank lower for long documents.

General chat at 8K context

  1. 89Excellent quality (A+) · 89 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 88/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 26B-A4B Q5_K_M · 3.7 tok/s · Slow (tight fit)
  2. 90Excellent quality (A+) · 90 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.8 27B Q4_K_M · 0.6 tok/s · Very slow (tight fit)
  3. 83Great quality (A) · 83 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Raon-1 mini Preview (30B-A4B) Q4_K_M · 4.0 tok/s · Slow (tight fit)
  4. 89Excellent quality (A+) · 89 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.6 27B Q4_K_M · 0.6 tok/s · Very slow (tight fit)
  5. 78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.6 35B-A3B Q3_K_M · 7.3 tok/s · Slow (tight fit)

Coding at 32K context

  1. 87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 88/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 26B-A4B Q4_K_M · 4.5 tok/s · Slow (tight fit)
  2. 83Great quality (A) · 83 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Raon-1 mini Preview (30B-A4B) Q4_K_M · 4.0 tok/s · Slow (tight fit)
  3. 78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-V 4.7 35B-A3B Q3_K_M · 7.3 tok/s · Slow (tight fit)
  4. 77Great quality (A) · 77 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Nex N2 Mini (35B-A3B) Q3_K_M · 7.3 tok/s · Slow (tight fit)
  5. 87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Muse Glimmer 30B Q4_K_M · 0.6 tok/s · Very slow (tight fit)

Reasoning at 32K context

  1. 83Great quality (A) · 83 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Raon-1 mini Preview (30B-A4B) Q4_K_M · 4.0 tok/s · Slow (tight fit)
  2. 78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-V 4.7 35B-A3B Q3_K_M · 7.3 tok/s · Slow (tight fit)
  3. 77Great quality (A) · 77 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Nex N2 Mini (35B-A3B) Q3_K_M · 7.3 tok/s · Slow (tight fit)
  4. 87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Muse Glimmer 30B Q4_K_M · 0.6 tok/s · Very slow (tight fit)
  5. 76Great quality (A) · 76 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. KAT-Coder V2.5 Dev (35B-A3B) Q3_K_M · 7.3 tok/s · Slow (tight fit)

Documents (RAG) at 32K context

  1. 87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 88/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 26B-A4B Q4_K_M · 4.5 tok/s · Slow (tight fit)
  2. 78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.6 35B-A3B Q3_K_M · 7.3 tok/s · Slow (tight fit)
  3. 78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-V 4.7 35B-A3B Q3_K_M · 7.3 tok/s · Slow (tight fit)
  4. 78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.5 35B-A3B Q3_K_M · 7.3 tok/s · Slow (tight fit)
  5. 87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Muse Glimmer 30B Q4_K_M · 0.6 tok/s · Very slow (tight fit)

Agents & tools at 64K context

  1. 83Great quality (A) · 83 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Raon-1 mini Preview (30B-A4B) Q4_K_M · 4.0 tok/s · Slow (tight fit)
  2. 78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-V 4.7 35B-A3B Q3_K_M · 7.3 tok/s · Slow (tight fit)
  3. 77Great quality (A) · 77 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Nex N2 Mini (35B-A3B) Q3_K_M · 7.3 tok/s · Slow (tight fit)
  4. 76Great quality (A) · 76 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. KAT-Coder V2.5 Dev (35B-A3B) Q3_K_M · 7.3 tok/s · Slow (tight fit)
  5. 77Great quality (A) · 77 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 26B-A4B Q3_K_M · 5.8 tok/s · Slow (tight fit)

Reverse engineering at 32K context

  1. 78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.5 35B-A3B Q3_K_M · 7.3 tok/s · Slow (tight fit)
  2. 73Great quality (A) · 73 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. GLM-4.7 Flash (30B-A3B) Q3_K_M · 7.3 tok/s · Slow (tight fit)
  3. 87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 26B-A4B Q4_K_M · 4.5 tok/s · Slow (tight fit)
  4. 70Great quality (A) · 70 of 100: how capable the model is at MXFP4, whatever it runs on. Match score 81/100 (quality, fit, speed and memory use) sets the ranking. gpt-oss 20B MXFP4 · 5.1 tok/s · Slow
  5. 83Great quality (A) · 83 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 81/100 (quality, fit, speed and memory use) sets the ranking. Raon-1 mini Preview (30B-A4B) Q4_K_M · 4.0 tok/s · Slow (tight fit)

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.

Was this page helpful?