Best local LLMs for the Rockchip RK3576 (8 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. 64Good quality (B) · 64 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking. Nanbeige4.2 3B Q4_K_M · 3.1 tok/s · Slow
  2. 64Good quality (B) · 64 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. Nemotron Labs Dory 1.0 4B Q4_K_M · 3.3 tok/s · Slow
  3. 61Good quality (B) · 61 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 E2B Q6_K · 4.0 tok/s · Slow
  4. 65Good quality (B) · 65 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 76/100 (quality, fit, speed and memory use) sets the ranking. SenseNova Loop 9B Preview Q3_K_M · 1.8 tok/s · Very slow (tight fit)
  5. 62Good quality (B) · 62 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 76/100 (quality, fit, speed and memory use) sets the ranking. Spark-X2.6 4B Q4_K_M · 3.2 tok/s · Slow

Coding at 32K context

  1. 64Good quality (B) · 64 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking. Nemotron Labs Dory 1.0 4B Q4_K_M · 3.3 tok/s · Slow
  2. 62Good quality (B) · 62 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. Spark-X2.6 4B Q4_K_M · 3.2 tok/s · Slow
  3. 59Good quality (B) · 59 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 76/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM5 2B Q6_K · 3.6 tok/s · Slow
  4. 60Good quality (B) · 60 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 E2B Q5_K_M · 4.7 tok/s · Slow
  5. 54Fair quality (C) · 54 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. MiMo V2.6 Distill Qwen 9B Q2_K · 2.3 tok/s · Very slow (tight fit)

Reasoning at 32K context

  1. 64Good quality (B) · 64 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking. Nemotron Labs Dory 1.0 4B Q4_K_M · 3.3 tok/s · Slow
  2. 62Good quality (B) · 62 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. Spark-X2.6 4B Q4_K_M · 3.2 tok/s · Slow
  3. 59Good quality (B) · 59 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 76/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM5 2B Q6_K · 3.6 tok/s · Slow
  4. 52Fair quality (C) · 52 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.5 4B Q3_K_M · 3.6 tok/s · Slow
  5. 60Good quality (B) · 60 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 E2B Q5_K_M · 4.7 tok/s · Slow

Documents (RAG) at 32K context

  1. 60Good quality (B) · 60 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 E2B Q5_K_M · 4.7 tok/s · Slow
  2. 57Good quality (B) · 57 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. TMax 4B Q4_K_M · 3.1 tok/s · Slow
  3. 58Good quality (B) · 58 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. Nemotron 3.5 Bolt E4B Q3_K_M · 3.8 tok/s · Slow (tight fit)
  4. 41Fair quality (C) · 41 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 74/100 (quality, fit, speed and memory use) sets the ranking. Unlimited OCR 2 (3B-A0.6B) Q4_K_M · 20.4 tok/s · Runs well
  5. 57Good quality (B) · 57 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 74/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 E4B Q3_K_M · 3.8 tok/s · Slow (tight fit)

Agents & tools at 64K context

  1. 62Good quality (B) · 62 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 76/100 (quality, fit, speed and memory use) sets the ranking. Spark-X2.6 4B Q4_K_M · 3.2 tok/s · Slow (tight fit)
  2. 59Good quality (B) · 59 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM5 2B Q6_K · 3.6 tok/s · Slow (tight fit)
  3. 58Good quality (B) · 58 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. Nemotron 3.5 Bolt E4B Q3_K_M · 3.8 tok/s · Slow (tight fit)
  4. 57Good quality (B) · 57 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 74/100 (quality, fit, speed and memory use) sets the ranking. TMax 4B Q4_K_M · 3.1 tok/s · Slow (tight fit)
  5. 60Good quality (B) · 60 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 E2B Q5_K_M · 4.7 tok/s · Slow

Reverse engineering at 32K context

  1. 64Good quality (B) · 64 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 73/100 (quality, fit, speed and memory use) sets the ranking. Nemotron Labs Dory 1.0 4B Q4_K_M · 3.3 tok/s · Slow
  2. 62Good quality (B) · 62 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. Spark-X2.6 4B Q4_K_M · 3.2 tok/s · Slow
  3. 60Good quality (B) · 60 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. Gemma 4 E2B Q5_K_M · 4.7 tok/s · Slow
  4. 59Good quality (B) · 59 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM5 2B Q6_K · 3.6 tok/s · Slow
  5. 57Good quality (B) · 57 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 70/100 (quality, fit, speed and memory use) sets the ranking. TMax 4B Q4_K_M · 3.1 tok/s · Slow

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?