Best Local LLMs for 24GB RAM

Ranked open-weight models that run well on a 24GB machine, with estimated speed and the run command for each.

Try: Qwen3 8B, RTX 4090, MacBook Pro M4, 16GB, iPhone 17 Pro

Ranked picks

Best local models for ~24GB

12 picks

Estimates assume a representative Apple-Silicon machine with 24GB unified memory. Tok/s are ModelFit estimates, not measured benchmarks; typical error is the median gap to measured runs on hardware the model was not fitted on (how the estimates work). Run the wizard for figures tuned to your exact chip. On a graphics card instead? Best LLM for 24GB VRAM.

  1. 01

    LFM2 24B-A2B Instruct

    LFM2 / 24B / Q4_K_M / ~14 GB

    This model may feel memory-heavy on 24 GB RAM, but it is still listed for balanced speed and quality.

    ollama
    $ollama run lfm2:24b-a2b
    Fit
    Runs wellC · 17
    Best for
    Local AI agents, privacy-first tool calling, MCP workflows
    Perf
    est. 109 tok/s, typical error 10% · first token ~0.2s
  2. 02

    Gemma 4 26B-A4B

    Gemma / 26B / Q4_K_M / ~16 GB

    This model may feel memory-heavy on 24 GB RAM, but it is still listed for balanced speed and quality.

    ollama
    $ollama run gemma4:26b
    Fit
    Runs wellC · 5
    Best for
    Chat, Coding, Multimodal
    Perf
    est. 63 tok/s, typical error 10% · first token ~0.4s
  3. 03

    Qwen3.5 9B Instruct (Q8)

    Qwen / 9B / Q8_0 / ~10.7 GB

    Best for quality, coding, reasoning. Strong fit for 24 GB RAM with balanced speed and quality.

    ollama
    $ollama run qwen3.5:9b-q8_0
    Fit
    Runs wellB · 36
    Best for
    Quality, Coding, Reasoning
    Perf
    est. 31 tok/s, typical error 10% · first token ~0.9s
  4. 04

    Qwen3.8 27B

    Qwen / 27B / Q4_K_M / ~16.5 GB

    This model may feel memory-heavy on 24 GB RAM, but it is still listed for balanced speed and quality.

    ollama
    $ollama run qwen3.8:27b
    Fit
    Runs wellC · 2
    Best for
    Coding, Agent, Vision, Long context
    Perf
    est. 17 tok/s, typical error 10% · first token ~2.6s
  5. 05

    GPT-OSS 20B

    GPT-OSS / 21B / MXFP4 / ~13.8 GB

    This model may feel memory-heavy on 24 GB RAM, but it is still listed for balanced speed and quality.

    ollama
    $ollama run gpt-oss:20b
    Fit
    Runs wellC · 18
    Best for
    Chat, Coding, Reasoning
    Perf
    est. 64 tok/s, typical error 10% · first token ~0.4s
  6. 06

    Qwen3.5 9B Instruct

    Qwen / 9B / Q4_K_M / ~7 GB

    Best for quality, coding, reasoning. Strong fit for 24 GB RAM with balanced speed and quality.

    ollama
    $ollama run qwen3.5:9b
    Fit
    Runs wellB · 58
    Best for
    Quality, Coding, Reasoning
    Perf
    est. 51 tok/s, typical error 10% · first token ~0.9s
  7. 07

    Gemma 4 12B (Q8)

    Gemma / 12B / Q8_0 / ~12.8 GB

    This model may feel memory-heavy on 24 GB RAM, but it is still listed for balanced speed and quality.

    ollama
    $ollama run gemma4:12b-it-q8_0
    Fit
    Runs wellC · 24
    Best for
    Chat, Coding, Multimodal
    Perf
    est. 20 tok/s, typical error 10% · first token ~1.3s
  8. 08

    Ministral 3 14B

    Mistral / 13.5B / Q4_K_M / ~9.1 GB

    Best for chat, reasoning, vision. Strong fit for 24 GB RAM with balanced speed and quality.

    ollama
    $ollama run ministral-3:14b
    Fit
    Runs wellB · 46
    Best for
    Chat, Reasoning, Vision
    Perf
    est. 29 tok/s, typical error 10% · first token ~1.5s
  9. 09

    Gemma 4 12B

    Gemma / 12B / Q4_K_M / ~8 GB

    Best for chat, coding, multimodal. Strong fit for 24 GB RAM with balanced speed and quality.

    ollama
    $ollama run gemma4:12b
    Fit
    Runs wellB · 52
    Best for
    Chat, Coding, Multimodal
    Perf
    est. 33 tok/s, typical error 10% · first token ~1.3s
  10. 10

    LFM2.5 8B-A1B

    LFM2 / 8.3B / Q4_K_M / ~5.5 GB

    Best for on-device agents, tool calling, multilingual chat. Strong fit for 24 GB RAM with balanced speed and quality.

    ollama
    $ollama run lfm2.5:8b-a1b-q4_K_M
    Fit
    Lots of roomA · 67
    Best for
    On-device agents, tool calling, multilingual chat
    Perf
    est. 142 tok/s, typical error 10% · first token ~0.2s
  11. 11

    NVIDIA Nemotron 3.5 Bolt E4B

    Nemotron / 4B / Q4_K_M / ~7.9 GB

    Best for voice assistants, tool calling, chat. Strong fit for 24 GB RAM with balanced speed and quality.

    ollama
    $ollama run nemotron-3.5-bolt:e4b
    Fit
    Runs wellB · 53
    Best for
    Voice assistants, Tool calling, Chat
    Perf
    est. 66 tok/s, typical error 10% · first token ~0.8s
  12. 12

    Mellum2.1 12B-A2.5B Thinking

    Mellum / 12.1B / Q4_K_M / ~8.1 GB

    Best for agentic coding, small machines. Strong fit for 24 GB RAM with balanced speed and quality.

    llama.cpp
    $llama-server -hf JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF:Q4_K_M
    Fit
    Runs wellB · 52
    Best for
    Agentic coding, Small machines
    Perf
    est. 98 tok/s, typical error 10% · first token ~0.3s

Browse by RAM tier

Want an exact recommendation?

The wizard tunes picks and speed estimates to your exact device, chip, and RAM.

Open ModelFit Wizard Browse all devices
Best Local LLMs for 24GB RAM (2026) | ModelFit