Find the machine for your model. Sized by memory.

51 model entries. Compare GPU memory figures and their sources before choosing a machine configuration.

These are catalog sizing thresholds, planning estimates and external reports. Context, concurrency and runtime overhead can increase memory use. A memory match does not establish a tested deployment, throughput or live capacity.

Sources and sizing assumptions

Kimi K2.6 MoE

• kimi-k2-6-moe · Profile runtime: ollama · Serve: 600 GB (4-bit) · System RAM: 768 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

DeepSeek V4 Pro

• deepseek-v4-pro · Profile runtime: ollama · Serve: 400 GB (4-bit) · System RAM: 512 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

MiniMax M2

• vllm-minimax-m2 · Profile runtime: vllm · Serve: 140 GB (Runtime default) · System RAM: 192 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vllm-minimax-m2 · Profile runtime: vllm · Serve: 138 GB (4-bit (NVFP4)) · System RAM: 64 GB

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• vllm-minimax-m2 · Profile runtime: vllm · QLoRA: 200 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

gpt-oss 120B

Read gpt-oss 120B requirements and sources

• gpt-oss-120b · Profile runtime: ollama · Serve: 80 GB (Runtime default) · System RAM: 128 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• gpt-oss-120b · Profile runtime: ollama · Serve: 72 GB (4-bit (MXFP4)) · System RAM: 64 GB

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• gpt-oss-120b · Profile runtime: ollama · QLoRA: 90 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• gpt-oss-120b · Profile runtime: ollama · LoRA: 360 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Llama 3.2 Vision 90B

• vision-llama32-90b · Profile runtime: ollama · Serve: 64 GB (Runtime default) · System RAM: 96 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llama32-90b · Profile runtime: ollama · Serve: 55 GB (4-bit) · System RAM: 64 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llama32-90b · Profile runtime: ollama · Serve: 216 GB (16-bit) · System RAM: 192 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llama32-90b · Profile runtime: ollama · QLoRA: 68 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llama32-90b · Profile runtime: ollama · LoRA: 270 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen2.5-VL 72B

• vision-qwen25vl-72b · Profile runtime: ollama · Serve: 48 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-qwen25vl-72b · Profile runtime: ollama · Serve: 45 GB (4-bit) · System RAM: 64 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-qwen25vl-72b · Profile runtime: ollama · Serve: 173 GB (16-bit) · System RAM: 192 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-qwen25vl-72b · Profile runtime: ollama · QLoRA: 54 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-qwen25vl-72b · Profile runtime: ollama · LoRA: 216 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Llama 3.3 70B

Read Llama 3.3 70B requirements and sources

• max-70b · Profile runtime: ollama · Serve: 80 GB (Runtime default) · System RAM: 128 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• max-70b · Profile runtime: ollama · Serve: 42 GB (4-bit (Q4_K_M)) · System RAM: 64 GB

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• max-70b · Profile runtime: ollama · Serve: 84 GB (8-bit (Q6_K_L)) · System RAM: 96 GB

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• max-70b · Profile runtime: ollama · Serve: 168 GB (16-bit) · System RAM: 192 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• max-70b · Profile runtime: ollama · QLoRA: 53 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• max-70b · Profile runtime: ollama · LoRA: 210 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

LLaVA 34B

• vision-llava-34b · Profile runtime: ollama · Serve: 24 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llava-34b · Profile runtime: ollama · QLoRA: 26 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llava-34b · Profile runtime: ollama · LoRA: 102 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen3 32B

Read Qwen3 32B requirements and sources

• quality-32b · Profile runtime: ollama · Serve: 32 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• quality-32b · Profile runtime: ollama · Serve: 22 GB (4-bit) · System RAM: 32 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• quality-32b · Profile runtime: ollama · Serve: 77 GB (16-bit) · System RAM: 96 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• quality-32b · Profile runtime: ollama · QLoRA: 24 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• quality-32b · Profile runtime: ollama · LoRA: 96 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen Coder 32B

• qwen-coder-32b · Profile runtime: ollama · Serve: 32 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen-coder-32b · Profile runtime: ollama · QLoRA: 24 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen-coder-32b · Profile runtime: ollama · LoRA: 96 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

DeepSeek-R1 32B

• deepseek-r1-32b · Profile runtime: ollama · Serve: 32 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• deepseek-r1-32b · Profile runtime: ollama · QLoRA: 24 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• deepseek-r1-32b · Profile runtime: ollama · LoRA: 96 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen2.5 32B Instruct

• qwen25-32b · Profile runtime: ollama · Serve: 32 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen25-32b · Profile runtime: ollama · QLoRA: 24 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen25-32b · Profile runtime: ollama · LoRA: 96 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen2.5-VL 32B

• vision-qwen25vl-32b · Profile runtime: ollama · Serve: 24 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-qwen25vl-32b · Profile runtime: ollama · QLoRA: 24 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-qwen25vl-32b · Profile runtime: ollama · LoRA: 96 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

MOVA 720p + audio (SGLang)

No sourced sizing row is recorded for this model.

Qwen3 30B A3B

• qwen3-30b · Profile runtime: ollama · Serve: 32 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-30b · Profile runtime: ollama · Serve: 20 GB (4-bit (Q4_K_M)) · System RAM: 32 GB

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• qwen3-30b · Profile runtime: ollama · QLoRA: 23 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-30b · Profile runtime: ollama · LoRA: 90 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-30b-a3b · Profile runtime: ollama · Serve: 32 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-30b-a3b · Profile runtime: ollama · QLoRA: 23 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-30b-a3b · Profile runtime: ollama · LoRA: 90 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen3 Coder 30B

Read Qwen3 Coder 30B requirements and sources

• qwen3-coder-30b · Profile runtime: ollama · Serve: 32 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-coder-30b · Profile runtime: ollama · Serve: 20 GB (4-bit (Q4_K_M)) · System RAM: 32 GB

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• qwen3-coder-30b · Profile runtime: ollama · QLoRA: 23 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-coder-30b · Profile runtime: ollama · LoRA: 90 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Gemma 3 27B

Read Gemma 3 27B requirements and sources

• gemma3-27b · Profile runtime: ollama · Serve: 24 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• gemma3-27b · Profile runtime: ollama · Serve: 18 GB (4-bit (Q4_K_M)) · System RAM: 32 GB

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• gemma3-27b · Profile runtime: ollama · Serve: 65 GB (16-bit) · System RAM: 64 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• gemma3-27b · Profile runtime: ollama · QLoRA: 21 GB (QLoRA)

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• gemma3-27b · Profile runtime: ollama · LoRA: 81 GB (LoRA)

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

Qwen3 Coder 27B

• qwen3-coder-27b · Profile runtime: ollama · Serve: 32 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-coder-27b · Profile runtime: ollama · QLoRA: 20 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-coder-27b · Profile runtime: ollama · LoRA: 81 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Mistral Small 24B

• mistral-small-24b · Profile runtime: ollama · Serve: 24 GB (Runtime default) · System RAM: 32 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• mistral-small-24b · Profile runtime: ollama · Serve: 16 GB (4-bit) · System RAM: 32 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• mistral-small-24b · Profile runtime: ollama · Serve: 29 GB (8-bit) · System RAM: 48 GB

  Model-card-derived planning figure. The row does not record a source URL; not a B3IQ measurement. Sizing note: VRAM includes 20% over the weights for context and runtime. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• mistral-small-24b · Profile runtime: ollama · QLoRA: 18 GB (QLoRA)

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• mistral-small-24b · Profile runtime: ollama · LoRA: 72 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Devstral Small 24B

• devstral-small-24b · Profile runtime: ollama · Serve: 16 GB (Runtime default) · System RAM: 32 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• devstral-small-24b · Profile runtime: ollama · QLoRA: 18 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• devstral-small-24b · Profile runtime: ollama · LoRA: 72 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Mistral Small 3.1 24B (Vision)

• vision-mistral-small31 · Profile runtime: ollama · Serve: 18 GB (Runtime default) · System RAM: 32 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-mistral-small31 · Profile runtime: ollama · QLoRA: 18 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-mistral-small31 · Profile runtime: ollama · LoRA: 72 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

gpt-oss 20B

Read gpt-oss 20B requirements and sources

• gpt-oss-20b · Profile runtime: ollama · Serve: 24 GB (Runtime default) · System RAM: 32 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• gpt-oss-20b · Profile runtime: ollama · Serve: 16 GB (4-bit) · System RAM: 32 GB

  External report. Not independently verified on B3IQ. The profile runtime may differ from the reported setup. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment. View source

• gpt-oss-20b · Profile runtime: ollama · QLoRA: 15 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• gpt-oss-20b · Profile runtime: ollama · LoRA: 60 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen3 14B

Read Qwen3 14B requirements and sources

• llamacpp-qwen3-14b · Profile runtime: llama.cpp · QLoRA: 11 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• llamacpp-qwen3-14b · Profile runtime: llama.cpp · LoRA: 42 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-14b · Profile runtime: ollama · Serve: 16 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-14b · Profile runtime: ollama · QLoRA: 11 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-14b · Profile runtime: ollama · LoRA: 42 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vllm-qwen3-14b · Profile runtime: vllm · Serve: 24 GB (Runtime default) · System RAM: 64 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vllm-qwen3-14b · Profile runtime: vllm · QLoRA: 11 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vllm-qwen3-14b · Profile runtime: vllm · LoRA: 42 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen2.5 Coder 14B

• code-14b · Profile runtime: ollama · Serve: 16 GB (Runtime default) · System RAM: 32 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• code-14b · Profile runtime: ollama · QLoRA: 11 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• code-14b · Profile runtime: ollama · LoRA: 42 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Phi-4 14B

• phi4-14b · Profile runtime: ollama · Serve: 16 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• phi4-14b · Profile runtime: ollama · QLoRA: 11 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• phi4-14b · Profile runtime: ollama · LoRA: 42 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

DeepSeek-R1 14B

• deepseek-r1-14b · Profile runtime: ollama · Serve: 16 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• deepseek-r1-14b · Profile runtime: ollama · QLoRA: 11 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• deepseek-r1-14b · Profile runtime: ollama · LoRA: 42 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen2.5 14B Instruct

• qwen25-14b · Profile runtime: ollama · Serve: 16 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen25-14b · Profile runtime: ollama · QLoRA: 11 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen25-14b · Profile runtime: ollama · LoRA: 42 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Wan2.1 14B

• video-wan21-14b-xinference · Profile runtime: xinference · Serve: 40 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

LLaVA 13B

• vision-llava-13b · Profile runtime: ollama · Serve: 10 GB (Runtime default) · System RAM: 16 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llava-13b · Profile runtime: ollama · QLoRA: 10 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llava-13b · Profile runtime: ollama · LoRA: 39 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

HunyuanVideo

• video-hunyuan-xinference · Profile runtime: xinference · Serve: 48 GB (Runtime default) · System RAM: 96 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Gemma 3 12B

• gemma3-12b · Profile runtime: ollama · Serve: 16 GB (Runtime default) · System RAM: 48 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• gemma3-12b · Profile runtime: ollama · QLoRA: 9 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• gemma3-12b · Profile runtime: ollama · LoRA: 36 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Llama 3.2 Vision 11B

• vision-llama32-11b · Profile runtime: ollama · Serve: 8 GB (Runtime default) · System RAM: 16 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llama32-11b · Profile runtime: ollama · QLoRA: 8 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llama32-11b · Profile runtime: ollama · LoRA: 33 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen3 8B

Read Qwen3 8B requirements and sources

• llamacpp-qwen3-8b · Profile runtime: llama.cpp · QLoRA: 6 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• llamacpp-qwen3-8b · Profile runtime: llama.cpp · LoRA: 24 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-8b · Profile runtime: ollama · QLoRA: 6 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-8b · Profile runtime: ollama · LoRA: 24 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Granite 3.3 8B

• granite33-8b · Profile runtime: ollama · QLoRA: 6 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• granite33-8b · Profile runtime: ollama · LoRA: 24 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Llama 3.1 8B

Read Llama 3.1 8B requirements and sources

• starter-8b · Profile runtime: ollama · Serve: 8 GB (Runtime default) · System RAM: 16 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• starter-8b · Profile runtime: ollama · QLoRA: 6 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• starter-8b · Profile runtime: ollama · LoRA: 24 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

DeepSeek-R1 8B

• deepseek-r1-8b · Profile runtime: ollama · Serve: 8 GB (Runtime default) · System RAM: 24 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• deepseek-r1-8b · Profile runtime: ollama · QLoRA: 6 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• deepseek-r1-8b · Profile runtime: ollama · LoRA: 24 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen3 Coder 8B

• qwen3-coder-8b · Profile runtime: ollama · QLoRA: 6 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen3-coder-8b · Profile runtime: ollama · LoRA: 24 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen3 Embedding 8B

• embed-qwen3-8b-xinference · Profile runtime: xinference · Serve: 18 GB (Runtime default) · System RAM: 32 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

MiniCPM-V 2.6 8B

• vision-minicpm-v · Profile runtime: ollama · Serve: 7 GB (Runtime default) · System RAM: 10 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-minicpm-v · Profile runtime: ollama · QLoRA: 6 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-minicpm-v · Profile runtime: ollama · LoRA: 24 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

LLaVA-Llama3 8B

• vision-llava-llama3 · Profile runtime: ollama · Serve: 7 GB (Runtime default) · System RAM: 10 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llava-llama3 · Profile runtime: ollama · QLoRA: 6 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llava-llama3 · Profile runtime: ollama · LoRA: 24 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Mistral 7B

• mistral-7b · Profile runtime: ollama · QLoRA: 5 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• mistral-7b · Profile runtime: ollama · LoRA: 21 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen Coder 7B

• qwen-coder-7b · Profile runtime: ollama · Serve: 8 GB (Runtime default) · System RAM: 24 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen-coder-7b · Profile runtime: ollama · QLoRA: 5 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen-coder-7b · Profile runtime: ollama · LoRA: 21 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

DeepSeek-R1 7B

• deepseek-r1-7b · Profile runtime: ollama · Serve: 8 GB (Runtime default) · System RAM: 24 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• deepseek-r1-7b · Profile runtime: ollama · QLoRA: 5 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• deepseek-r1-7b · Profile runtime: ollama · LoRA: 21 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen2.5 7B Instruct

• qwen25-7b · Profile runtime: ollama · Serve: 8 GB (Runtime default) · System RAM: 24 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen25-7b · Profile runtime: ollama · QLoRA: 5 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• qwen25-7b · Profile runtime: ollama · LoRA: 21 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

LLaVA 7B (Vision)

• vision-llava-7b · Profile runtime: ollama · QLoRA: 5 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-llava-7b · Profile runtime: ollama · LoRA: 21 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

gte-Qwen2 7B

• embed-gte-qwen2-xinference · Profile runtime: xinference · Serve: 16 GB (Runtime default) · System RAM: 32 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Qwen2.5-VL 7B

Read Qwen2.5-VL 7B requirements and sources

• vision-qwen25vl-7b · Profile runtime: ollama · Serve: 7 GB (Runtime default) · System RAM: 10 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-qwen25vl-7b · Profile runtime: ollama · QLoRA: 5 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-qwen25vl-7b · Profile runtime: ollama · LoRA: 21 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

BakLLaVA 7B

• vision-bakllava · Profile runtime: ollama · Serve: 6 GB (Runtime default) · System RAM: 8 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-bakllava · Profile runtime: ollama · QLoRA: 5 GB (QLoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

• vision-bakllava · Profile runtime: ollama · LoRA: 21 GB (LoRA)

  Planning estimate. Not a B3IQ measurement. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

BGE-M3

Read BGE-M3 requirements and sources

No sourced sizing row is recorded for this model.

Whisper Large v3 STT (LocalAI)

Read Whisper large-v3 requirements and sources

No sourced sizing row is recorded for this model.

Whisper Large v3 STT

Read Whisper large-v3 requirements and sources

• audio-whisper-v3-xinference · Profile runtime: xinference · Serve: 6 GB (Runtime default) · System RAM: 8 GB

  Catalog threshold. Configuration guidance, not measured peak memory. No row-specific sizing note is recorded. Test date, measured context length, concurrency, runtime version and test hardware are not recorded for this row. Confirm these before deployment.

Hardware · Machines