Best local LLMs for the Rockchip RK3576 (4 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. 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. MiniCPM5 2B Q4_K_M · 5.2 tok/s · Slow (tight fit)
  2. 46Fair quality (C) · 46 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 70/100 (quality, fit, speed and memory use) sets the ranking. Spark-X2.6 4B Q2_K · 5.2 tok/s · Slow (tight fit)
  3. 39Basic quality (D) · 39 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 69/100 (quality, fit, speed and memory use) sets the ranking. LFM3 1.7B Q2_K · 12.5 tok/s · Runs well
  4. 42Fair quality (C) · 42 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking. TMax 4B Q2_K · 5.1 tok/s · Slow (tight fit)
  5. 41Fair quality (C) · 41 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking. LFM2.5 2.6B Q4_K_M · 4.9 tok/s · Slow (tight fit)

Coding at 32K context

  1. 34Basic quality (D) · 34 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM5 1B Q4_K_M · 12.1 tok/s · Runs well
  2. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. PaddleOCR-VL 1.6 0.9B Q4_K_M · 13.7 tok/s · Runs well
  3. 39Basic quality (D) · 39 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. LFM3 1.7B Q2_K · 12.5 tok/s · Runs well
  4. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 65/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-OCR 1.0 1.3B Q4_K_M · 10.1 tok/s · Runs well
  5. 35Basic quality (D) · 35 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking. LFM3-VL 2B Q2_K · 11.2 tok/s · Runs well

Reasoning at 32K context

  1. 34Basic quality (D) · 34 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM5 1B Q4_K_M · 12.1 tok/s · Runs well
  2. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. PaddleOCR-VL 1.6 0.9B Q4_K_M · 13.7 tok/s · Runs well
  3. 36Basic quality (D) · 36 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. Jamba Reasoning 3B Q3_K_M · 5.3 tok/s · Slow (tight fit)
  4. 39Basic quality (D) · 39 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. LFM3 1.7B Q2_K · 12.5 tok/s · Runs well
  5. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 65/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-OCR 1.0 1.3B Q4_K_M · 10.1 tok/s · Runs well

Documents (RAG) at 32K context

  1. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. PaddleOCR-VL 1.6 0.9B Q4_K_M · 13.7 tok/s · Runs well
  2. 39Basic quality (D) · 39 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. LFM3 1.7B Q2_K · 12.5 tok/s · Runs well
  3. 39Basic quality (D) · 39 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. MiniCPM-OCR 1.0 1.3B Q4_K_M · 10.1 tok/s · Runs well
  4. 35Basic quality (D) · 35 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking. LFM3-VL 2B Q2_K · 11.2 tok/s · Runs well
  5. 33Basic quality (D) · 33 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 67/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-V 4.7 1B Q3_K_M · 10.9 tok/s · Runs well

Agents & tools at 64K context

  1. 39Basic quality (D) · 39 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. LFM3 1.7B Q2_K · 12.5 tok/s · Runs well
  2. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 65/100 (quality, fit, speed and memory use) sets the ranking. PaddleOCR-VL 1.6 0.9B Q4_K_M · 13.7 tok/s · Runs well (tight fit)
  3. 28Basic quality (D) · 28 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 64/100 (quality, fit, speed and memory use) sets the ranking. TMax 2B Q2_K · 11.5 tok/s · Runs well (tight fit)
  4. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 64/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-OCR 1.0 1.3B Q4_K_M · 10.1 tok/s · Runs well (tight fit)
  5. 35Basic quality (D) · 35 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 62/100 (quality, fit, speed and memory use) sets the ranking. LFM3-VL 2B Q2_K · 11.2 tok/s · Runs well (tight fit)

Reverse engineering at 32K context

  1. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. PaddleOCR-VL 1.6 0.9B Q4_K_M · 13.7 tok/s · Runs well
  2. 39Basic quality (D) · 39 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. LFM3 1.7B Q2_K · 12.5 tok/s · Runs well
  3. 39Basic quality (D) · 39 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 65/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM-OCR 1.0 1.3B Q4_K_M · 10.1 tok/s · Runs well
  4. 35Basic quality (D) · 35 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking. LFM3-VL 2B Q2_K · 11.2 tok/s · Runs well
  5. 34Basic quality (D) · 34 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking. MiniCPM5 1B Q4_K_M · 12.1 tok/s · Runs well

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?