What Models?

Pick the right model for your GPU — in seconds.

Best local LLMs for 8 GB VRAM (November 2026): what fits, context and speed

With 8 GB of VRAM or unified memory, 33 models run well and 4 more are a tight fit. 70 don't fit.

Speeds assume 320 GB/s, the median memory bandwidth of the 33 GPUs and memory configurations with exactly 8 GB. Yours may be faster or slower: pick your GPU for exact numbers.

Best overall

Highest MMLU-Pro score among models that run well

  1. Qwen3.5 9B Q4_K_M 87K context · ~48 tok/s · MMLU 82.5
  2. Qwen3.5 4B Q8_0 123K context · ~58 tok/s · MMLU 79.1
  3. Qwen3 8B Q4_K_M 24K context · ~53 tok/s · MMLU 79
  4. Gemma 4 12B Q4_K_M 91K context · ~39 tok/s · MMLU 77.2
  5. Nemotron Nano 12B Q4_K_M 44K context · ~38 tok/s · MMLU 76

Best for coding

Highest SWE-bench Verified score among models that run well

  1. K2-Horizon 7B Q4_K_M 20K context · ~49 tok/s · SWE-bench 72.4
  2. Ornith 1.5 9B Q4_K_M 84K context · ~47 tok/s · SWE-bench 70.6
  3. MiMo V2.6 Distill Qwen 9B Q4_K_M 82K context · ~47 tok/s · SWE-bench 61.1
  4. Mellum2.1 12B A2.5B Thinking Q4_K_M 36K context · ~92 tok/s · SWE-bench 47
  5. LFM3 1.7B Q8_0 256K context · ~112 tok/s · SWE-bench 43.4

GPUs with 8 GB of memory

Graphics cards

Unified memory (Apple Silicon, Ryzen AI Max, RTX Spark)

  • M2 100 GB/s
  • M3 100 GB/s
  • M1 68 GB/s

Upgrade options

12 GB runs 9 more models well, including Phi-4 14B, Qwen 2.5 14B and GPT-OSS 20B.

The cheapest GPUs with more than 8 GB, by launch MSRP.

As an Amazon Associate, I earn from qualifying purchases.

Every model in 8 GB

61 of 214 model configurations can run on 8 GB — sorted by quality

107 models, each counted once by the best way it runs on your hardware.

Largest that runs well Nemotron Nano 12B 12.6B params · Q4_K_M · 44K context
Fastest good-quality model Phi-4 Mini 3.8B ~92 tok/s · Q4_K_M · 46K context
? MMLU (Massive Multitask Language Understanding) measures general knowledge across 57 subjects. Higher = more capable. Sorted best-first.

Runs well 33

  • Qwen3.5 9B 9B Great VisionTool useReasoning
    MMLU 82.5
    Q4_K_M
    Context 87K Speed ~48 tok/s
    Q8_0 Doesn't fit · needs 8.9 GB
  • Qwen3.5 4B 4B Great VisionTool useReasoning
    MMLU 79.1
    Q4_K_M
    Context 175K Speed ~86 tok/s
    Q8_0
    Context 123K Speed ~58 tok/s
  • Qwen3 8B 8.2B Great ReasoningTool use
    MMLU 79
    Q4_K_M
    Context 24K Speed ~53 tok/s
    Q8_0 Doesn't fit · needs 8.2 GB

Runs well 33

  • Qwen3.5 9B 9B Great VisionTool useReasoning
    MMLU 82.5
    Q4_K_M
    Context 87K Speed ~48 tok/s
    Q8_0 Doesn't fit · needs 8.9 GB
  • Qwen3.5 4B 4B Great VisionTool useReasoning
    MMLU 79.1
    Q4_K_M
    Context 175K Speed ~86 tok/s
    Q8_0
    Context 123K Speed ~58 tok/s
  • Qwen3 8B 8.2B Great ReasoningTool use
    MMLU 79
    Q4_K_M
    Context 24K Speed ~53 tok/s
    Q8_0 Doesn't fit · needs 8.2 GB
  • Gemma 4 12B 11.96B Great VisionTool useReasoning
    MMLU 77.2
    Q4_K_M
    Context 91K Speed ~39 tok/s
    Q8_0 Doesn't fit · needs 11.8 GB
  • Nemotron Nano 12B 12.6B Great ReasoningTool use
    MMLU 76
    Q4_K_M
    Context 44K Speed ~38 tok/s
    Q8_0 Doesn't fit · needs 12.2 GB
  • Qwen 2.5 7B 7.62B Good Tool use
    MMLU 74.2
    Q4_K_M
    Context 68K Speed ~56 tok/s
    Q8_0 Tight fit · 8K context
  • Qwen 2.5 VL 7B 7.62B Good VisionTool use
    MMLU 74.2
    Q4_K_M
    Context 68K Speed ~56 tok/s
    Q8_0 Tight fit · 8K context
  • Gemma 3 12B 11.8B Good Vision
    MMLU 74
    Q4_K_M
    Context 19K Speed ~39 tok/s
    Q8_0 Doesn't fit · needs 11.7 GB
  • Llama 3.2 Vision 11B 10.6B Good VisionTool use
    MMLU 73
    Q4_K_M
    Context 16K Speed ~43 tok/s
    Q8_0 Doesn't fit · needs 10.5 GB
  • Phi-4 Mini 3.8B 3.84B Good
    MMLU 70
    Q4_K_M
    Context 46K Speed ~92 tok/s
    Q8_0
    Context 34K Speed ~63 tok/s
  • Gemma 4 E4B 4.5B Good VisionTool useReasoning
    MMLU 69.4
    Q4_K_M
    Context 128K Speed ~50 tok/s
    Q8_0
    Context 34K Speed ~35 tok/s
  • Llama 3.1 8B 8.03B Good Tool use
    MMLU 69.4
    Q4_K_M
    Context 28K Speed ~54 tok/s
    Q8_0 Doesn't fit · needs 8.1 GB
    fp16 Doesn't fit · needs 15.1 GB
  • Qwen 2.5 Coder 7B 7.62B Good
    MMLU 68
    Q4_K_M
    Context 68K Speed ~56 tok/s
    Q8_0 Tight fit · 8K context
  • Qwen 2.5 VL 3B 3.09B Fair Vision
    MMLU 65.6
    Q4_K_M
    Context 128K Speed ~109 tok/s
    Q8_0
    Context 128K Speed ~75 tok/s
  • Qwen 2.5 3B 3.09B Fair Tool use
    MMLU 65.6
    Q4_K_M
    Context 32K Speed ~109 tok/s
    Q8_0
    Context 32K Speed ~75 tok/s
  • SmallThinker 3B 3.09B Fair Reasoning
    MMLU 65
    Q4_K_M
    Context 32K Speed ~103 tok/s
    Q8_0
    Context 32K Speed ~71 tok/s
  • Ministral 8B 8.02B Fair Tool use
    MMLU 65
    Q4_K_M
    Context 21K Speed ~50 tok/s
    Q8_0 Doesn't fit · needs 8.1 GB
  • Llama 3.2 3B 3.21B Fair Tool use
    MMLU 63.4
    Q4_K_M
    Context 57K Speed ~106 tok/s
    Q8_0
    Context 44K Speed ~72 tok/s
    fp16
    Context 18K Speed ~43 tok/s
  • Gemma 3 4B 3.88B Fair Vision
    MMLU 63
    Q4_K_M
    Context 128K Speed ~92 tok/s
    Q8_0
    Context 128K Speed ~62 tok/s
  • Gemma 4 E2B 2.3B Fair VisionTool useReasoning
    MMLU 60
    Q4_K_M
    Context 128K Speed ~72 tok/s
    Q8_0
    Context 128K Speed ~54 tok/s
  • DeepSeek R1 Distill Qwen 7B 7.62B Fair Reasoning
    MMLU 58
    Q4_K_M
    Context 68K Speed ~56 tok/s
    Q8_0 Tight fit · 8K context
  • Qwen3.5 2B 2B Fair VisionTool useReasoning
    MMLU 55.3
    Q4_K_M
    Context 256K Speed ~140 tok/s
    Q8_0
    Context 256K Speed ~106 tok/s
  • Llama 3.2 1B 1.24B Basic Tool use
    MMLU 49.3
    Q4_K_M
    Context 128K Speed ~177 tok/s
    Q8_0
    Context 128K Speed ~138 tok/s
  • Qwen3.5 0.8B 0.8B Basic VisionTool useReasoning
    MMLU 29.7
    Q4_K_M
    Context 256K Speed ~208 tok/s
    Q8_0
    Context 256K Speed ~176 tok/s
  • Nemotron 3 Nano 4B 3.97B Basic ReasoningTool use
    MMLU
    Q4_K_M
    Context 256K Speed ~81 tok/s
    Q8_0
    Context 256K Speed ~61 tok/s
  • LFM3 1.7B 1.73B Basic Tool useReasoning
    MMLU
    Q4_K_M
    Context 256K Speed ~153 tok/s
    Q8_0
    Context 256K Speed ~112 tok/s
  • LFM2.5 2.6B 2.7B Basic Tool use
    MMLU
    Q4_K_M
    Context 128K Speed ~120 tok/s
    Q8_0
    Context 128K Speed ~83 tok/s
  • LFM3-VL 2B 1.92B Basic VisionTool useReasoning
    MMLU
    Q4_K_M
    Context 128K Speed ~153 tok/s
    Q8_0
    Context 128K Speed ~112 tok/s
  • Ornith 1.5 9B 9.65B Basic Tool useReasoning
    MMLU
    Q4_K_M
    Context 84K Speed ~47 tok/s
    Q8_0 Doesn't fit · needs 9.1 GB
  • MiMo V2.6 Distill Qwen 9B 9.41B Basic VisionTool useReasoning
    MMLU
    Q4_K_M
    Context 82K Speed ~47 tok/s
    Q8_0 Doesn't fit · needs 8.9 GB
  • Nemotron 3.5 Bolt E4B 8B Basic Tool useReasoning
    MMLU
    Q4_K_M
    Context 64K Speed ~45 tok/s
    Q8_0
    Context 34K Speed ~35 tok/s
  • Mellum2.1 12B A2.5B Thinking 12.15B Basic Tool useReasoning
    MMLU
    Q4_K_M
    Context 36K Speed ~92 tok/s

    MoE — speed based on ~1.55 GB active experts plus per-token routing overhead (full 7.52 GB loaded in VRAM)

    Q8_0 Doesn't fit · needs 12.1 GB
  • K2-Horizon 7B 9B Basic Tool useReasoning
    MMLU
    Q4_K_M
    Context 20K Speed ~49 tok/s
    Q8_0 Doesn't fit · needs 9.1 GB

Tight fit 4

  • Gemma 2 9B 9.24B Good
    MMLU 71.3
    Q4_K_M
    Context 8K Speed ~47 tok/s

    Only 8K context — under the 16K needed to be practical

    Q8_0 Doesn't fit · needs 9.5 GB
  • Phi-3 Mini 3.8B 3.82B Good
    MMLU 70.9
    Q4_K_M
    Context 15K Speed ~94 tok/s

    Only 15K context — under the 16K needed to be practical

    Q8_0
    Context 11K Speed ~63 tok/s

    Only 11K context — under the 16K needed to be practical

    GPUs that can run this

    As an Amazon Associate, I earn from qualifying purchases.

  • Mistral Nemo 12B 12.2B Good Tool use
    MMLU 68
    Q4_K_M
    Context 5K Speed ~38 tok/s

    Only 5K context — under the 16K needed to be practical

    Q8_0 Doesn't fit · needs 12.3 GB
    GPUs that can run this

    As an Amazon Associate, I earn from qualifying purchases.

Tight fit 4

  • Gemma 2 9B 9.24B Good
    MMLU 71.3
    Q4_K_M
    Context 8K Speed ~47 tok/s

    Only 8K context — under the 16K needed to be practical

    Q8_0 Doesn't fit · needs 9.5 GB
  • Phi-3 Mini 3.8B 3.82B Good
    MMLU 70.9
    Q4_K_M
    Context 15K Speed ~94 tok/s

    Only 15K context — under the 16K needed to be practical

    Q8_0
    Context 11K Speed ~63 tok/s

    Only 11K context — under the 16K needed to be practical

    GPUs that can run this

    As an Amazon Associate, I earn from qualifying purchases.

  • Mistral Nemo 12B 12.2B Good Tool use
    MMLU 68
    Q4_K_M
    Context 5K Speed ~38 tok/s

    Only 5K context — under the 16K needed to be practical

    Q8_0 Doesn't fit · needs 12.3 GB
    GPUs that can run this

    As an Amazon Associate, I earn from qualifying purchases.

  • DeepSeek Coder 6.7B 6.74B Basic
    MMLU 50
    Q4_K_M
    Context 8K Speed ~63 tok/s

    Only 8K context — under the 16K needed to be practical

    Q8_0
    Context 2K Speed ~39 tok/s

    Only 2K context — under the 16K needed to be practical

    GPUs that can run this

    As an Amazon Associate, I earn from qualifying purchases.

Notify me when new models fit my GPU

An email when a new model fits your 8 GB setup — no spam, unsubscribe anytime.

Doesn't fit 70

  • DeepSeek R1 0528 671B Excellent Reasoning
    MMLU 90.8
    Q4_K_M Needs 381.2 GB — 373.2 GB over
    Q8_0 Needs 664.4 GB — 656.4 GB over
  • Qwen3 Next 80B-A3B 80B Excellent Tool use
    MMLU 90
    Q4_K_M Needs 45.1 GB — 37.1 GB over
    Q8_0 Needs 79.0 GB — 71.0 GB over
    GPUs that can run this

    As an Amazon Associate, I earn from qualifying purchases.

  • MiMo V2.5 Pro 1020B Excellent Reasoning
    MMLU 89.4
    Q4_K_M Needs 586.5 GB — 578.5 GB over
    Q8_0 Needs 1013.0 GB — 1005.0 GB over

Doesn't fit 70