DeskGPU
English

NVIDIA GeForce RTX 4070: which LLMs can it run?

This page was written and translated automatically by AI and checked by software against its sources, with no human editing. Specifications come from the sources listed; memory requirements and speeds are estimates calculated with published formulas. How we use AI

Specifications are for the 2023 version (AD104-250, GDDR6X); other versions sold under the same name are not covered.

NVIDIA GeForce RTX 4070 has 12 GB of memory (manufacturer figure). Below, every model in the database is checked against it with llama.cpp at a context length of 8,192 tokens (or the model’s maximum if shorter).

Values updated Oct 19, 2026

Specifications

Memory (manufacturer figure)
12 GB
Memory type
GDDR6X
Memory bus width
192 bit
Memory bandwidth
504 GB/s
Board power (TDP)
200 W
FP16 tensor, dense
Not shown: open sources do not give two independent, agreeing values with a manufacturer reference.
BF16 tensor, dense
Not shown: open sources do not give two independent, agreeing values with a manufacturer reference.
FP8 tensor, dense
Not shown: open sources do not give two independent, agreeing values with a manufacturer reference.
FP4 tensor, dense
Not shown: open sources do not give two independent, agreeing values with a manufacturer reference.
INT8 tensor, dense
Not shown: open sources do not give two independent, agreeing values with a manufacturer reference.
Interface
Not shown: open sources do not give two independent, agreeing values with a manufacturer reference.

Each value lists every source with its own figure; units have been normalized. A manufacturer’s “GB” of GPU memory is counted as GiB (2³⁰ bytes), which is also how AMD’s ROCm table states it. How specifications are checked

Run profile

Run profile: llama.cpp, CUDA back end (NVIDIA discrete GPU), all layers on the GPU, Flash Attention on, micro-batch (-ub) 512, F16 KV cache, context (-c) as shown; “Fits” is judged with llama-server’s default 4 parallel sequences (-np auto, unified KV cache) and “Doesn’t fit” with 1 sequence; idle GPU with no other programs using its memory.

Memory counted as available for “Fits” (C_low): 8.88 GiB per card = 12 GiB nominal − 1,054 MiB (the largest gap between nominal and reported total measured on consumer and workstation GPUs in this profile, used because no public log reports this GPU’s own total) − 2,112 MiB (the largest start-up and runtime use seen in this profile’s public logs), rounded down to a multiple of 64 MiB.

These results assume the card’s default configuration. Settings that reduce usable memory (for example ECC, where the card supports it) are not covered.

Memory used for “Doesn’t fit” (C_high): the full 12 GiB per card. Run profiles and their measurements

Which models fit

“Fits” means llama.cpp's default --fit load check would place all layers on the GPU: our estimate of llama.cpp's buffers, plus its error margin, plus llama.cpp's 1 GiB per-card reserve, is within the memory counted as available. It does not promise that peak memory while generating fits. How results are decided

Context 8,192 tokens (or the model’s maximum if shorter). The last columns show the longest context at that quantization and the bandwidth-based speed ceiling.
Model Parameters Highest quantization that fits Available memory needed for “Fits” Longest context Speed ceiling Result
Trinity-Mini 26.1B (3.1B active) IQ2_XXS 8.2 GiB 41,216 ≤ 367 tok/s Fits
Trinity-Mini-Base 26.1B (3.1B active) IQ2_XXS 8.7 GiB 16,640 ≤ 405 tok/s Fits
Trinity-Mini-Base-Pre-Anneal 26.1B (3.1B active) IQ2_XXS 8.4 GiB 4,096 ≤ 428 tok/s Fits
Qwen3-Coder-REAP-25B-A3B 24.9B (3.0B active) IQ2_XXS 8.1 GiB 15,872 ≤ 269 tok/s Fits
DeepSeek-R1-Distill-Qwen-14B 14.8B Q2_K 8.3 GiB 11,008 ≤ 70.8 tok/s Fits
Qwen2.5-14B-Instruct 14.8B Q2_K 8.3 GiB 11,008 ≤ 70.8 tok/s Fits
ContextPilot-14B 14.8B IQ3_XXS 8.4 GiB 10,752 ≤ 69.5 tok/s Fits
FrogMini-14B-2510 14.8B IQ3_XXS 8.1 GiB 12,544 ≤ 72.6 tok/s Fits
JanusCoder-14B 14.8B Q2_K 8.0 GiB 13,312 ≤ 73.7 tok/s Fits
Nemotron-Cascade-14B-Thinking 14.8B IQ3_XXS 8.1 GiB 12,544 ≤ 72.6 tok/s Fits
Nemotron-Terminal-14B 14.8B IQ3_XXS 8.1 GiB 12,544 ≤ 72.6 tok/s Fits
NousCoder-14B 14.8B IQ3_XXS 8.1 GiB 12,544 ≤ 72.6 tok/s Fits
Qwen3-Nemotron-14B-BRRM 14.8B IQ3_XXS 8.1 GiB 12,544 ≤ 72.6 tok/s Fits
WebWorld-14B 14.8B IQ3_XXS 8.1 GiB 12,544 ≤ 72.6 tok/s Fits
phi-4 14.7B Q2_K 8.3 GiB 11,008 ≤ 70.8 tok/s Fits
Apriel-1.6-15b-Thinker 14.4B IQ3_XXS 8.2 GiB 11,264 ≤ 71.0 tok/s Fits
Mistral-Nemo-Instruct-2407 12.2B IQ4_XS 8.8 GiB 8,192 ≤ 65.3 tok/s Fits
Mellum2-12B-A2.5B-Base 12.1B (2.2B active) Q4_0 8.3 GiB 45,824 ≤ 340 tok/s Fits
Mellum2-12B-A2.5B-Instruct 12.1B (2.2B active) Q4_0 8.3 GiB 41,728 ≤ 325 tok/s Fits
Mellum2-12B-A2.5B-Instruct-SFT 12.1B (2.2B active) Q4_0 8.3 GiB 45,824 ≤ 340 tok/s Fits
Mellum2-12B-A2.5B-Thinking 12.1B (2.2B active) Q4_0 8.3 GiB 41,728 ≤ 325 tok/s Fits
Mellum2-12B-A2.5B-Thinking-SFT 12.1B (2.2B active) Q4_0 8.3 GiB 45,824 ≤ 340 tok/s Fits
gemma-4-12B-it 11.9B Q3_K_M 8.7 GiB 16,128 ≤ 82.0 tok/s Fits
gemma-3-12b-it 11.8B Q2_K 8.2 GiB 17,920 ≤ 89.5 tok/s Fits
translategemma-12b-it 11.8B Q2_K 8.2 GiB 17,920 ≤ 89.5 tok/s Fits
AutoGLM-Phone-9B 9.4B Q5_K_M 8.1 GiB 25,600 ≤ 72.5 tok/s Fits
AutoGLM-Phone-9B-Multilingual 9.4B Q5_K_M 8.1 GiB 25,600 ≤ 72.5 tok/s Fits
GLM-4.6V-Flash 9.4B Q5_K_M 8.1 GiB 25,600 ≤ 72.5 tok/s Fits
gemma-2-9b-it 9.2B IQ3_XXS 8.0 GiB 8,192 ≤ 85.4 tok/s Fits
EuroLLM-9B-Instruct-2512 9.2B Q5_K_M 8.7 GiB 9,216 ≤ 66.6 tok/s Fits
Fara1.5-9B 9.0B Q6_K 8.7 GiB 13,056 ≤ 70.8 tok/s Fits
Holo-3.1-9B 9.0B Q6_K 8.4 GiB 22,784 ≤ 74.3 tok/s Fits
MiMo-V2.6-Distill-Qwen-9B 9.0B Q6_K 8.8 GiB 10,240 ≤ 69.8 tok/s Fits
Qwen3.5-9B 9.0B Q6_K 8.5 GiB 19,968 ≤ 73.2 tok/s Fits
Qwen3.5-9B-Base 9.0B Q6_K 8.4 GiB 22,784 ≤ 74.3 tok/s Fits
UI-Mate-9B 9.0B Q6_K 8.7 GiB 13,056 ≤ 70.8 tok/s Fits
UI-Venus-2-9B 9.0B Q6_K 8.7 GiB 13,056 ≤ 70.8 tok/s Fits
NVIDIA-Nemotron-Nano-9B-v2 8.9B Q5_K_M 8.5 GiB 30,208 ≤ 74.6 tok/s Fits
granite-4.1-8b 8.8B Q5_K_M 8.4 GiB 10,752 ≤ 68.9 tok/s Fits
granite-4.2-8b 8.8B Q5_K_M 8.4 GiB 10,752 ≤ 68.9 tok/s Fits
Ministral-3-8B-Instruct-2512 8.5B Q5_K_M 8.0 GiB 14,592 ≤ 73.9 tok/s Fits
Ministral-3-8B-Reasoning-2512 8.5B Q5_K_M 8.0 GiB 14,592 ≤ 73.9 tok/s Fits
LFM2.5-8B-A1B 8.5B (1.7B active) Q6_K 8.2 GiB 56,576 ≤ 339 tok/s Fits
granite-guardian-4.1-8b 8.4B Q5_K_M 8.4 GiB 10,752 ≤ 68.9 tok/s Fits
LFM2-8B-A1B 8.3B (1.6B active) Q6_K 8.1 GiB 63,744 ≤ 364 tok/s Fits
rnj-1-instruct 8.3B Q5_K_M 8.2 GiB 13,568 ≤ 71.9 tok/s Fits
AesCode-8B 8.2B Q5_K_M 8.2 GiB 12,544 ≤ 71.8 tok/s Fits
ArmorOCR 8.2B Q4_K_M 7.1 GiB 20,224 ≤ 85.7 tok/s Fits
AstaBrief_8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
ContextPilot-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Cosmos-Reason2-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
DeepSeek-R1-0528-Qwen3-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Holo2-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
intent-aware-lfqa-qwen3-8b-baseline 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
intent-aware-lfqa-qwen3-8b-intent-explicit 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
intent-aware-lfqa-qwen3-8b-intent-implicit 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
intent-aware-lfqa-qwen3-8b-multiview 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
JanusCoder-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
MathForm-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Nemotron-Cascade-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Nemotron-Cascade-8B-Thinking 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Nemotron-Orchestrator-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Nemotron-Terminal-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Qwen3-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Qwen3-Nemotron-8B-BRRM 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Qwen3-VL-8B-Instruct 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Qwen3-VL-8B-Thinking 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
RLVR-8B-0926 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
Step3-VL-10B-Base 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
UI-Venus-1.5-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
WebWorld-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
ZwZ-8B 8.2B Q6_K 8.6 GiB 9,984 ≤ 67.9 tok/s Fits
MiniCPM-o-4_5 8.2B Q6_K 8.6 GiB 9,984 ≤ 68.0 tok/s Fits
granite-3.3-8b-instruct 8.2B Q5_K_M 8.3 GiB 11,776 ≤ 70.6 tok/s Fits
granite-guardian-3.2-8b-factuality-detection 8.2B Q5_K_M 8.3 GiB 11,776 ≤ 70.6 tok/s Fits
intent-aware-lfqa-llama3-8b-baseline 8.0B Q6_K 8.4 GiB 12,032 ≤ 69.7 tok/s Fits
intent-aware-lfqa-llama3-8b-intent-explicit 8.0B Q6_K 8.4 GiB 12,032 ≤ 69.7 tok/s Fits
intent-aware-lfqa-llama3-8b-intent-implicit 8.0B Q6_K 8.4 GiB 12,032 ≤ 69.7 tok/s Fits
intent-aware-lfqa-llama3-8b-multiview 8.0B Q6_K 8.4 GiB 12,032 ≤ 69.7 tok/s Fits
DeepSeek-R1-Distill-Llama-8B 8.0B Q6_K 8.4 GiB 12,032 ≤ 69.7 tok/s Fits
HARC-Llama-3.1-8B-Instruct 8.0B Q6_K 8.4 GiB 12,032 ≤ 69.7 tok/s Fits
Llama-3.1-8B-Instruct 8.0B Q6_K 8.4 GiB 12,032 ≤ 69.7 tok/s Fits
AprielGuard 7.9B Q6_K 7.9 GiB 18,176 ≤ 75.0 tok/s Fits
MiMo-Embodied-7B 7.6B Q6_K 8.1 GiB 13,056 ≤ 72.5 tok/s Fits
DeepSeek-R1-Distill-Qwen-7B 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
Fara-7B 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
HARC-Qwen2.5-7B-Instruct 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
JanusCoderV-7B 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
olmOCR-2-7B-1025 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
Qwen2.5-7B-Instruct 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
Qwen2.5-Coder-7B-Instruct 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
Qwen2.5-VL-7B-Instruct 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
Qwen2.5-VL-7B-Surg-CholecT50 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
Spatial-SSRL-7B 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
X-Reasoner-7B 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
ZwZ-7B 7.6B Q6_K 7.5 GiB 32,256 ≤ 80.4 tok/s Fits
gemma-4-E4B 7.5B Q8_0 6.7 GiB 128,256 ≤ 97.4 tok/s Fits
gemma-4-E4B-it 7.5B Q8_0 6.9 GiB 119,296 ≤ 94.4 tok/s Fits
Bielik-Minitron-7B-v3.0-Instruct 7.5B Q6_K 8.5 GiB 10,240 ≤ 68.4 tok/s Fits
ContextPilot-E4B 7.5B Q8_0 6.6 GiB 131,072 ≤ 98.5 tok/s Fits
Olmo-3.1-7B-RL-Zero-Code 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3.1-7B-RL-Zero-Math 7.3B Q2_K 8.2 GiB 9,472 ≤ 93.3 tok/s Fits
Olmo-3-7B-Instruct 7.3B Q2_K 8.2 GiB 9,472 ≤ 93.3 tok/s Fits
Olmo-3-7B-Instruct-DPO 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3-7B-Instruct-SFT 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3-7B-RL-Zero-Code 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3-7B-RL-Zero-General 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3-7B-RL-Zero-IF 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3-7B-RL-Zero-Math 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3-7B-RL-Zero-Mix 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3-7B-Think 7.3B Q2_K 8.2 GiB 9,472 ≤ 93.3 tok/s Fits
Olmo-3-7B-Think-DPO 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Olmo-3-7B-Think-SFT 7.3B IQ3_XXS 8.2 GiB 9,472 ≤ 93.2 tok/s Fits
Mistral-7B-Instruct-v0.2 7.2B Q6_K 8.1 GiB 14,336 ≤ 73.0 tok/s Fits
Trinity-Nano-Base 6.1B (819M active) Q6_K 6.7 GiB 131,072 ≤ 560 tok/s Fits
Trinity-Nano-Base-Pre-Anneal 6.1B (819M active) Q6_K 6.4 GiB 4,096 ≤ 599 tok/s Fits
Trinity-Nano-Preview 6.1B (819M active) Q8_0 8.1 GiB 56,320 ≤ 459 tok/s Fits
Youtu-VL-4B-Instruct 4.9B Q6_K 5.9 GiB 32,768 ≤ 114 tok/s Fits
Holo-3.1-4B 4.8B Q8_0 6.4 GiB 81,408 ≤ 106 tok/s Fits
gemma-4-E2B 4.6B Q6_K 3.5 GiB 131,072 ≤ 257 tok/s Fits
gemma-4-E2B-it 4.6B BF16 6.1 GiB 131,072 ≤ 108 tok/s Fits
AgentCPM-Explore 4.4B Q6_K 5.7 GiB 29,440 ≤ 112 tok/s Fits
CapRL-Qwen3VL-4B 4.4B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
Holo2-4B 4.4B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
intent-aware-lfqa-qwen3-4b-baseline 4.4B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
intent-aware-lfqa-qwen3-4b-intent-explicit 4.4B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
intent-aware-lfqa-qwen3-4b-intent-implicit 4.4B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
intent-aware-lfqa-qwen3-4b-multiview 4.4B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
Spatial-SSRL-Qwen3VL-4B 4.4B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
ZwZ-4B 4.4B Q6_K 5.7 GiB 29,440 ≤ 112 tok/s Fits
Fara1.5-4B 4.2B Q8_0 6.4 GiB 81,152 ≤ 106 tok/s Fits
FrogNano-4B-2609 4.2B Q8_0 6.4 GiB 81,152 ≤ 106 tok/s Fits
Qwen3.5-4B 4.2B Q8_0 6.4 GiB 81,408 ≤ 106 tok/s Fits
Qwen3.5-4B-Base 4.2B Q8_0 6.4 GiB 81,408 ≤ 106 tok/s Fits
UserRepGen-4B 4.2B Q8_0 6.4 GiB 81,152 ≤ 106 tok/s Fits
Riva-Translate-4B-Instruct-v1.1 4.2B Q6_K 6.0 GiB 8,192 ≤ 105 tok/s Fits
GELab-Zero-4B-preview 4.0B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
Qwen3-4B 4.0B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
Qwen3-4B-Instruct-2507 4.0B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
Qwen3-VL-4B-Instruct 4.0B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
Qwen3-VL-4B-Thinking 4.0B Q8_0 6.7 GiB 23,040 ≤ 91.9 tok/s Fits
NVIDIA-Nemotron-3-Nano-4B-BF16 4.0B BF16 8.7 GiB 15,872 ≤ 69.4 tok/s Fits
gemma-3-4b-it 3.9B Q8_0 6.3 GiB 124,928 ≤ 114 tok/s Fits
medgemma-1.5-4b-it 3.9B Q8_0 6.3 GiB 124,928 ≤ 114 tok/s Fits
Nemotron-3-Content-Safety 3.9B Q6_K 5.4 GiB 131,072 ≤ 145 tok/s Fits
translategemma-4b-it 3.9B Q8_0 6.3 GiB 124,928 ≤ 114 tok/s Fits
Phi-4-mini-instruct 3.8B Q8_0 6.5 GiB 26,624 ≤ 97.9 tok/s Fits
Apertus-v1.1-4B 3.8B Q6_K 4.4 GiB 4,096 ≤ 157 tok/s Fits
Phi-3.5-mini-instruct 3.8B Q8_0 8.3 GiB 9,472 ≤ 70.2 tok/s Fits
granite-4.2-3b 3.7B BF16 8.6 GiB 11,264 ≤ 67.4 tok/s Fits
BitCPM-CANN-3B-unquantized 3.6B Q6_K 4.3 GiB 32,768 ≤ 164 tok/s Fits
kanana-2-3b-instruct 3.5B F16 8.6 GiB 10,496 ≤ 67.8 tok/s Fits
Ministral-3-3B-Instruct-2512 3.4B BF16 8.9 GiB 8,192 ≤ 65.2 tok/s Fits
Ministral-3-3B-Reasoning-2512 3.4B BF16 8.9 GiB 8,192 ≤ 65.2 tok/s Fits
granite-vision-4.1-4b 3.4B BF16 8.6 GiB 11,264 ≤ 67.4 tok/s Fits
granite-4.1-3b 3.4B BF16 8.6 GiB 11,264 ≤ 67.4 tok/s Fits
granite-guardian-4.0-3b-toxicity-ja 3.4B Q6_K 4.7 GiB 59,136 ≤ 146 tok/s Fits
Qwen2.5-3B-Instruct 3.4B Q8_0 4.8 GiB 32,768 ≤ 141 tok/s Fits
tiny-aya-earth 3.3B BF16 8.7 GiB 10,240 ≤ 71.2 tok/s Fits
tiny-aya-fire 3.3B BF16 8.7 GiB 10,240 ≤ 71.2 tok/s Fits
tiny-aya-global 3.3B BF16 8.7 GiB 10,240 ≤ 71.2 tok/s Fits
tiny-aya-water 3.3B BF16 8.7 GiB 10,240 ≤ 71.2 tok/s Fits
Llama-3.2-3B-Instruct 3.2B BF16 8.5 GiB 11,264 ≤ 68.4 tok/s Fits
AI21-Jamba-Reasoning-3B 3.2B BF16 7.3 GiB 262,144 ≤ 83.0 tok/s Fits
AI21-Jamba2-3B 3.2B BF16 7.3 GiB 262,144 ≤ 83.0 tok/s Fits
kanana-2-3b-base 3.2B Q6_K 4.9 GiB 32,768 ≤ 137 tok/s Fits
SmolLM3-3B 3.1B BF16 7.9 GiB 21,248 ≤ 74.6 tok/s Fits
SmolLM3-3B-GSM8K-SFT 3.1B Q6_K 4.3 GiB 65,536 ≤ 161 tok/s Fits
LFM2-2.6B-Exp 2.7B F16 6.5 GiB 128,000 ≤ 95.6 tok/s Fits
d1-3B 2.7B BF16 6.7 GiB 32,768 ≤ 91.2 tok/s Fits
LFM2.5-2.6B 2.7B BF16 6.7 GiB 128,256 ≤ 91.2 tok/s Fits
LFM2.5-2.6B-Base 2.7B Q6_K 3.6 GiB 128,000 ≤ 215 tok/s Fits
LFM2.5-VL-3B 2.7B BF16 6.7 GiB 32,768 ≤ 91.2 tok/s Fits
EuroMoE-2.6B-A0.6B-Instruct-2512 2.6B (498M active) Q6_K 3.6 GiB 32,768 ≤ 820 tok/s Fits
LFM2-2.6B-Longevity 2.6B F16 6.5 GiB 128,000 ≤ 95.6 tok/s Fits
LFM2-2.6B-Transcript 2.6B BF16 6.5 GiB 128,000 ≤ 95.6 tok/s Fits
LFM2-VL-3B 2.6B BF16 6.5 GiB 128,000 ≤ 95.6 tok/s Fits
MiniCPM5-2B 2.5B BF16 6.0 GiB 72,448 ≤ 104 tok/s Fits
MiniCPM5-2B-Base 2.5B Q6_K 3.4 GiB 131,328 ≤ 229 tok/s Fits
CapRL-Qwen3VL-2B 2.0B F16 5.6 GiB 36,608 ≤ 115 tok/s Fits
HiPO-1.7B 2.0B Q6_K 3.6 GiB 40,960 ≤ 214 tok/s Fits
UI-Venus-1.5-2B 2.0B F16 5.6 GiB 36,608 ≤ 115 tok/s Fits
ZwZ-2B 2.0B F16 5.6 GiB 36,608 ≤ 115 tok/s Fits
Youtu-LLM-2B 2.0B Q6_K 3.2 GiB 131,072 ≤ 263 tok/s Fits
Youtu-LLM-2B-Base 2.0B Q6_K 3.2 GiB 131,072 ≤ 263 tok/s Fits
Qwen3.5-2B 1.9B BF16 5.4 GiB 256,512 ≤ 130 tok/s Fits
Qwen3.5-2B-Base 1.9B Q6_K 3.2 GiB 262,144 ≤ 306 tok/s Fits
Qwen3-1.7B 1.7B BF16 5.6 GiB 36,608 ≤ 115 tok/s Fits
Qwen3-VL-2B-Instruct 1.7B BF16 5.6 GiB 36,608 ≤ 115 tok/s Fits
Qwen3-VL-2B-Thinking 1.7B BF16 5.6 GiB 36,608 ≤ 115 tok/s Fits
SingGuard-2b 1.7B F16 5.6 GiB 36,608 ≤ 115 tok/s Fits
granite-4.0-1b 1.6B BF16 5.1 GiB 53,504 ≤ 128 tok/s Fits
Apertus-v1.1-1.5B 1.5B F16 3.8 GiB 4,096 ≤ 193 tok/s Fits
Apertus-v1.1-1.5B-Instruct 1.5B F16 3.8 GiB 4,096 ≤ 193 tok/s Fits
granite-4.0-h-1b 1.5B BF16 4.4 GiB 131,072 ≤ 168 tok/s Fits
Llama-3.2-1B-Instruct 1.2B BF16 4.0 GiB 131,072 ≤ 184 tok/s Fits
LFM2.5-1.2B-Base 1.2B BF16 3.7 GiB 128,000 ≤ 206 tok/s Fits
LFM2.5-1.2B-Instruct 1.2B BF16 3.7 GiB 128,000 ≤ 206 tok/s Fits
LFM2.5-1.2B-JP 1.2B F16 3.7 GiB 128,000 ≤ 206 tok/s Fits
LFM2.5-1.2B-JP-202606 1.2B F16 3.7 GiB 128,000 ≤ 206 tok/s Fits
LFM2.5-1.2B-Thinking 1.2B BF16 3.7 GiB 128,000 ≤ 206 tok/s Fits
LFM2.5-VL-1.6B 1.2B BF16 3.7 GiB 128,000 ≤ 206 tok/s Fits
LFM2.5-VL-1.6B-Extract 1.2B F16 3.7 GiB 128,000 ≤ 206 tok/s Fits
MiniCPM5-1B 1.1B Q6_K 2.2 GiB 131,072 ≤ 546 tok/s Fits
MiniCPM5-1B-Base 1.1B Q6_K 2.2 GiB 131,072 ≤ 546 tok/s Fits
MiniCPM5-1B-SFT 1.1B Q6_K 2.2 GiB 131,072 ≤ 546 tok/s Fits
Holo-3.1-0.8B 1.0B Q6_K 2.3 GiB 262,144 ≤ 701 tok/s Fits
MiniCPM-V-4.7-1B 1.0B F16 3.2 GiB 262,144 ≤ 314 tok/s Fits
gemma-3-1b-it 1.0B BF16 3.6 GiB 32,768 ≤ 246 tok/s Fits
Qwen3.5-0.8B 752M BF16 3.2 GiB 262,144 ≤ 314 tok/s Fits
Qwen3.5-0.8B-Base 752M Q6_K 2.3 GiB 262,144 ≤ 701 tok/s Fits
MiniCPM-V-4.6 752M F16 3.2 GiB 262,144 ≤ 314 tok/s Fits
MiniCPM-V-4.6-Thinking 752M F16 3.2 GiB 262,144 ≤ 314 tok/s Fits
Qwen3-0.6B 752M BF16 3.4 GiB 40,960 ≤ 236 tok/s Fits
HunyuanOCR 539M F16 3.1 GiB 32,768 ≤ 268 tok/s Fits
LFM2.5-350M 354M BF16 2.1 GiB 128,000 ≤ 622 tok/s Fits
LFM2.5-Encoder-350M-Diffusion 354M Q6_K 1.7 GiB 128,000 ≤ 1,287 tok/s Fits
LFM2.5-VL-450M 354M BF16 2.1 GiB 128,000 ≤ 622 tok/s Fits
LFM2.5-VL-450M-Extract 354M F16 2.1 GiB 128,000 ≤ 622 tok/s Fits
granite-4.0-350m 352M BF16 2.2 GiB 32,768 ≤ 536 tok/s Fits
granite-4.0-h-350m 340M BF16 2.1 GiB 32,768 ≤ 705 tok/s Fits
functiongemma-270m-it 268M BF16 2.1 GiB 32,768 ≤ 885 tok/s Fits
LFM2.5-230M 230M BF16 1.8 GiB 128,000 ≤ 900 tok/s Fits

Borderline models

Within the estimate error range: it may fit or it may not. Try a shorter context or a lower quantization.

Context 8,192 tokens. No quantization of these models fits; the one shown is the highest that is borderline.
Model Parameters Highest borderline quantization Available memory needed for “Fits” Result
AI21-Jamba2-Mini 51.6B (11.8B active) IQ2_XXS 13.6 GiB Tight
Kimi-Linear-48B-A3B-Base 49.1B (3.1B active) IQ2_XXS 14.2 GiB Tight
Kimi-Linear-48B-A3B-Instruct 49.1B (3.1B active) IQ2_XXS 13.3 GiB Tight
Hermes-4.3-36B 36.2B IQ2_XXS 12.9 GiB Tight
Kimi-Linear-REAP-35B-A3B-Instruct 35.1B (3.1B active) IQ3_XXS 14.8 GiB Tight
Holo-3.1-35B-A3B 34.7B (2.9B active) Q2_K 14.4 GiB Tight
Holo3-35B-A3B 34.7B (2.9B active) Q2_K 14.4 GiB Tight
Holo4-35B-A3B 34.7B (2.9B active) Q2_K 15.0 GiB Tight
Ising-Calibration-1-35B-A3B 34.7B (2.9B active) IQ3_XXS 15.1 GiB Tight
KAT-Coder-V2.5-Dev 34.7B (2.9B active) Q2_K 14.1 GiB Tight
Qwen3.5-35B-A3B-Base 34.7B (2.9B active) IQ3_XXS 15.0 GiB Tight
Qwen3.6-35B-A3B 34.7B (2.9B active) Q2_K 14.1 GiB Tight
Laguna-XS-2.1 33.4B (2.8B active) Q2_K 13.8 GiB Tight
AesCode-32B 32.8B IQ2_XXS 13.6 GiB Tight
Cosmos-Reason2-32B 32.8B IQ2_M 14.2 GiB Tight
FrogBoss-32B-2510 32.8B IQ2_M 14.2 GiB Tight
Nemotron-Terminal-32B 32.8B IQ2_M 14.2 GiB Tight
SERA-32B 32.8B IQ2_M 14.2 GiB Tight
WebWorld-32B 32.8B IQ2_M 14.2 GiB Tight
Olmo-3.1-32B-Instruct 32.2B Q2_K 15.0 GiB Tight
Olmo-3.1-32B-Instruct-DPO 32.2B IQ3_XXS 15.5 GiB Tight
Olmo-3.1-32B-Think 32.2B Q2_K 15.0 GiB Tight
Olmo-3-1125-32B 32.2B IQ3_XXS 15.5 GiB Tight
Olmo-3-32B-Think 32.2B Q2_K 15.0 GiB Tight
Olmo-3-32B-Think-DPO 32.2B IQ3_XXS 15.5 GiB Tight
Olmo-3-32B-Think-SFT 32.2B IQ3_XXS 15.5 GiB Tight
Ising-Calibration-1.5-31B-BF16 30.7B IQ2_XXS 16.5 GiB Tight
Holo2-30B-A3B 30.5B (3.0B active) Q2_K 13.0 GiB Tight
Qwen3-30B-A3B-Instruct-2507 30.5B (3.0B active) Q2_K 13.0 GiB Tight
Qwen3-Coder-30B-A3B-Instruct 30.5B (3.0B active) Q2_K 13.0 GiB Tight
Qwen3-VL-30B-A3B-Instruct 30.5B (3.0B active) Q2_K 13.0 GiB Tight
UI-Venus-1.5-30B-A3B 30.5B (3.0B active) Q2_K 13.0 GiB Tight
North-Mini-Code-1.0 30.5B (3.3B active) IQ3_XXS 15.1 GiB Tight
GLM-4.7-Flash 29.9B (3.6B active) Q2_K 12.8 GiB Tight
granite-4.2-30b 29.3B Q2_K 13.9 GiB Tight
granite-4.1-30b 28.9B Q2_K 13.9 GiB Tight
Muse-Glimmer-30B 27.9B Q3_K_M 15.0 GiB Tight
gemma-2-27b-it 27.2B IQ3_XXS 15.1 GiB Tight
gemma-3-27b-it 27.0B Q2_K 14.4 GiB Tight
translategemma-27b-it 27.0B Q2_K 14.4 GiB Tight
Fara1.5-27B 26.9B IQ3_XXS 14.3 GiB Tight
Holo4-27B 26.9B Q3_K_M 15.2 GiB Tight
Qwen3.5-27B 26.9B Q3_K_M 15.3 GiB Tight
Qwen3.8-27B 26.9B Q3_K_M 14.9 GiB Tight
Qwen3.8-3.6-27B-blend 26.9B Q3_K_M 14.9 GiB Tight
UI-Mate-27B 26.9B IQ3_XXS 14.3 GiB Tight
UI-Mate-democua-27B 26.9B Q3_K_M 15.2 GiB Tight
gemma-4-26B-A4B 25.2B (3.8B active) Q2_K 13.0 GiB Tight
gemma-4-26B-A4B-it 25.2B (3.8B active) Q3_K_M 15.4 GiB Tight
LFM2.5-24B-A2B 24.0B (2.5B active) Q4_0 14.7 GiB Tight
LFM2-24B-A2B 23.8B (2.3B active) Q4_0 14.6 GiB Tight
Devstral-Small-2-24B-Instruct-2512 23.6B IQ4_XS 14.7 GiB Tight
Mistral-Small-3.2-24B-Instruct-2506 23.6B IQ4_XS 14.7 GiB Tight
GLM-4.7-Flash-REAP-23B-A3B 23.0B (3.6B active) Q4_0 14.6 GiB Tight
EuroLLM-22B-2512 22.6B IQ4_XS 14.7 GiB Tight
EuroLLM-22B-Instruct-2512 22.6B IQ4_XS 14.7 GiB Tight
gpt-oss-20b 20.9B (3.6B active) Q8_0 12.9 GiB Tight
OptiMind-SFT 20.9B (3.6B active) Q4_0 13.2 GiB Tight
Qwen3-14B 14.8B Q6_K 13.9 GiB Tight
Ministral-3-14B-Instruct-2512 13.5B Q5_K_M 11.6 GiB Tight
Ministral-3-14B-Reasoning-2512 13.5B Q5_K_M 11.6 GiB Tight
Mellum2.1-12B-A2.5B-Thinking 12.1B (2.2B active) Q8_0 14.0 GiB Tight
gemma-4-12B 11.9B Q8_0 15.6 GiB Tight
Bielik-11B-v3.0-Instruct 11.2B Q8_0 14.4 GiB Tight

Models that don’t fit on one NVIDIA GeForce RTX 4070

Context 8,192 tokens. Even the lowest quantization does not fit on one card; the requirement shown is for that quantization.
Model Parameters Smallest quantization Available memory needed for “Fits” Several identical cards Result
DeepSeek-R1-Distill-Qwen-32B 32.8B Q2_K 15.2 GiB 2 identical cards (default even split) are expected to fit. Doesn't fit
Qwen2.5-32B-Instruct 32.8B Q2_K 15.2 GiB 2 identical cards (default even split) are expected to fit. Doesn't fit
Qwen2.5-Coder-32B-Instruct 32.8B Q2_K 15.2 GiB 2 identical cards (default even split) are expected to fit. Doesn't fit
Qwen3-VL-32B-Instruct 32.8B Q2_K 15.3 GiB 2 identical cards (default even split) are expected to fit. Doesn't fit
Qwen3-VL-32B-Thinking 32.8B Q2_K 15.3 GiB 2 identical cards (default even split) are expected to fit. Doesn't fit
Qwen3.6-27B 26.9B Q3_K_M 15.4 GiB 2 identical cards (default even split) are expected to fit. Doesn't fit
Qwen3.5-35B-A3B 34.7B (2.9B active) Q3_K_M 17.6 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
gemma-4-31B 30.7B Q2_K 17.6 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Holotron4-30B-A3B 31.6B (3.2B active) IQ2_M 18.9 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
NVIDIA-Nemotron-Labs-3-Elastic-30B-A3B-BF16 31.6B (3.2B active) IQ2_M 18.9 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Base-BF16 31.6B (3.2B active) Q2_K 18.9 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Nemotron-Cascade-2-30B-A3B 31.6B (3.2B active) IQ2_M 19.1 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 31.6B (3.2B active) IQ2_M 19.1 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 31.6B (3.2B active) IQ4_XS 19.1 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
MiniCPM-V-4.7-35B-A3B 34.7B (2.9B active) IQ4_XS 20.0 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Qwen3-30B-A3B 30.5B (3.0B active) Q4_K_M 20.1 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
gemma-4-31B-it 30.7B Q3_K_M 20.3 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
EXAONE-4.5-33B 32.5B IQ4_XS 20.4 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Llama-3.1-70B-Instruct 70.6B IQ2_XXS 22.3 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Qwen3-Coder-Next-Base 79.7B (3.6B active) IQ2_XXS 22.3 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Qwen3-32B 32.8B Q4_K_M 22.4 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
QwQ-32B 32.8B Q4_K_M 22.5 GiB 3 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 31.6B (3.2B active) Q4_K_M 23.3 GiB 4 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
KAT-Dev-72B-Exp 72.7B IQ2_XXS 28.5 GiB 4 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Solar-Open-100B 103B (11.2B active) IQ2_XXS 29.4 GiB 5 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
DeepSeek-R1-Distill-Llama-70B 70.6B Q2_K 29.4 GiB 4 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Llama-3.3-70B-Instruct 70.6B Q2_K 29.4 GiB 4 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Qwen3-Next-80B-A3B-Instruct 79.7B (3.6B active) Q2_K 30.3 GiB 5 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Qwen2.5-72B-Instruct 73.0B Q2_K 30.3 GiB 5 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Qwen3-Coder-Next 79.7B (3.6B active) Q2_K 30.4 GiB 5 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
GLM-4.5-Air-REAP-82B-A12B 81.9B (12.8B active) IQ2_XXS 31.1 GiB 5 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Step-3.5-Flash-REAP-121B-A11B 121B (11.4B active) IQ2_XXS 32.9 GiB 5 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Mistral-Small-4-119B-2603 119B (6.1B active) IQ2_XXS 34.3 GiB 7 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Qwen3.5-122B-A10B 122B (9.0B active) IQ2_XXS 35.0 GiB 6 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Qwen-AgentWorld-35B-A3B 34.7B (2.9B active) Q8_0 37.6 GiB 6 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
MiniMax-M2.5-REAP-139B-A10B 139B (10.4B active) IQ2_XXS 39.2 GiB 6 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
INTELLECT-3 107B (12.8B active) IQ2_XXS 39.6 GiB 6 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Step-3.5-Flash-REAP-149B-A11B 149B (11.4B active) IQ2_XXS 40.1 GiB 6 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Llama-4-Scout-17B-16E-Instruct 108B (16.1B active) Q2_K 41.5 GiB 6 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
GLM-4.5-Air 107B (12.8B active) Q2_K 45.6 GiB 7 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
GLM-4.6V 107B (12.8B active) Q2_K 45.6 GiB 7 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Step-3.5-Flash 197B (11.5B active) IQ2_XXS 48.9 GiB 8 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Devstral-2-123B-Instruct-2512 125B Q2_K 49.3 GiB 7 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
Mistral-Medium-3.5-128B 125B Q2_K 49.3 GiB 7 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16 121B (12.2B active) IQ2_XXS 49.8 GiB 7 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
NVIDIA-Nemotron-3-Super-120B-A12B-BF16 121B (12.2B active) IQ2_XXS 51.0 GiB 7 identical cards (default even split) are expected to fit. Fewer cards are borderline. Doesn't fit
gpt-oss-120b 117B (5.1B active) Q2_K 62.8 GiB Borderline with 8 cards. Doesn't fit
MiniMax-M2.7 229B (10.4B active) IQ2_XXS 62.9 GiB Borderline with 8 cards. Doesn't fit
Baichuan-M3-235B 235B (21.6B active) IQ2_XXS 63.2 GiB Borderline with 8 cards. Doesn't fit
Holo2-235B-A22B 235B (21.6B active) IQ2_XXS 63.2 GiB Borderline with 8 cards. Doesn't fit
Qwen3-Nemotron-235B-A22B-GenRM 235B (21.6B active) IQ2_XXS 63.2 GiB Borderline with 8 cards. Doesn't fit
command-a-plus-05-2026-bf16 218B (25.0B active) IQ2_XXS 63.6 GiB Borderline with 8 cards. Doesn't fit
NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 121B (12.2B active) Q4_0 71.8 GiB Borderline with 8 cards. Doesn't fit
GLM-4.7-REAP-268B-A32B 269B (32.8B active) IQ2_XXS 73.0 GiB Borderline with 8 cards. Doesn't fit
Step-3.7-Flash 197B (11.5B active) IQ3_XXS 76.1 GiB Borderline with 8 cards. Doesn't fit
Hy3 295B (20.2B active) IQ2_XXS 82.2 GiB Borderline with 8 cards. Doesn't fit
MiMo-V2.5 309B (14.8B active) IQ2_XXS 82.4 GiB Borderline with 8 cards. Doesn't fit
GLM-4.7-REAP-218B-A32B 218B (32.8B active) Q2_K 82.9 GiB Borderline with 8 cards. Doesn't fit
MiniMax-M2 229B (10.4B active) Q2_K 84.9 GiB Borderline with 8 cards. Doesn't fit
MiniMax-M2.1 229B (10.4B active) Q2_K 84.9 GiB Borderline with 8 cards. Doesn't fit
MiniMax-M2.5 229B (10.4B active) Q2_K 84.9 GiB Borderline with 8 cards. Doesn't fit
Qwen3-235B-A22B-Instruct-2507 235B (21.6B active) Q2_K 86.7 GiB Borderline with 8 cards. Doesn't fit
Qwen3-VL-235B-A22B-Instruct 235B (21.6B active) Q2_K 86.7 GiB Borderline with 8 cards. Doesn't fit
GLM-5.3-Flash-BF16 313B (16.7B active) IQ2_XXS 92.0 GiB Borderline with 8 cards. Doesn't fit
Laguna-S-2.1 118B (8.1B active) Q4_K_M 95.9 GiB Borderline with 8 cards. Doesn't fit
Trinity-Large-Base 399B (12.8B active) IQ2_XXS 96.4 GiB Borderline with 8 cards. Doesn't fit
Trinity-Large-TrueBase 399B (12.8B active) IQ2_XXS 96.4 GiB Borderline with 8 cards. Doesn't fit
GLM-4.6-REAP-268B-A32B 269B (32.8B active) Q2_K 100.9 GiB Borderline with 8 cards. Doesn't fit
Trinity-Large-Thinking 399B (12.8B active) IQ2_XXS 107.1 GiB Borderline with 8 cards. Doesn't fit
MiMo-V2-Flash 309B (14.8B active) Q2_K 111.6 GiB Does not fit even on 8 cards. Doesn't fit
Intern-S2-397B 396B (16.3B active) IQ2_XXS 114.3 GiB Does not fit even on 8 cards. Doesn't fit
MiMo-V2.6-Flash-MOPD 309B (14.8B active) Q2_K 124.9 GiB Does not fit even on 8 cards. Doesn't fit
MiMo-V2.6-Flash-RL 309B (14.8B active) Q2_K 124.9 GiB Does not fit even on 8 cards. Doesn't fit
K-EXAONE-236B-A23B 237B (22.7B active) IQ4_XS 126.3 GiB Does not fit even on 8 cards. Doesn't fit
GLM-4.5 353B (32.9B active) Q2_K 130.7 GiB Does not fit even on 8 cards. Doesn't fit
GLM-4.7 353B (32.9B active) Q2_K 130.7 GiB Does not fit even on 8 cards. Doesn't fit
Qwen3-Coder-REAP-363B-A35B 363B (34.5B active) Q2_K 132.7 GiB Does not fit even on 8 cards. Doesn't fit
Qwen3-235B-A22B 235B (21.6B active) Q4_K_M 141.9 GiB Does not fit even on 8 cards. Doesn't fit
Qwen3-Coder-480B-A35B-Instruct 480B (34.5B active) Q2_K 174.3 GiB Does not fit even on 8 cards. Doesn't fit
Qwen3.5-397B-A17B 396B (16.3B active) Q3_K_M 175.6 GiB Does not fit even on 8 cards. Doesn't fit
GLM-5.1 744B (40.8B active) IQ2_XXS 197.6 GiB Does not fit even on 8 cards. Doesn't fit
DeepSeek-V3.2 671B (36.6B active) Q2_K 242.1 GiB Does not fit even on 8 cards. Doesn't fit
cogito-671b-v2.1 671B (36.6B active) Q2_K 242.4 GiB Does not fit even on 8 cards. Doesn't fit
DeepSeek-V3.1 671B (36.6B active) Q2_K 242.4 GiB Does not fit even on 8 cards. Doesn't fit
Mistral-Large-3-675B-Instruct-2512 673B (39.0B active) Q2_K 243.8 GiB Does not fit even on 8 cards. Doesn't fit
DeepSeek-V3 671B (36.6B active) Q2_K 257.8 GiB Does not fit even on 8 cards. Doesn't fit
Kimi-K2.6 1.03T (31.7B active) IQ2_XXS 267.6 GiB Does not fit even on 8 cards. Doesn't fit
GLM-5 744B (40.8B active) Q2_K 268.9 GiB Does not fit even on 8 cards. Doesn't fit
MiMo-V2.5-Pro 1.02T (41.0B active) IQ2_XXS 270.4 GiB Does not fit even on 8 cards. Doesn't fit
DeepSeek-R1 671B (36.6B active) Q2_K 280.3 GiB Does not fit even on 8 cards. Doesn't fit
Muse-Spark-Open-816B-A42B 820B (44.4B active) Q2_K 301.0 GiB Does not fit even on 8 cards. Doesn't fit
GLM-5.3-Flash 313B (16.7B active) Q8_0 327.9 GiB Does not fit even on 8 cards. Doesn't fit
Ling-1T 1.00T (49.8B active) Q2_K 360.6 GiB Does not fit even on 8 cards. Doesn't fit
Kimi-K2-Instruct 1.03T (31.7B active) Q2_K 367.4 GiB Does not fit even on 8 cards. Doesn't fit
Kimi-K2.5 1.03T (31.7B active) Q2_K 368.0 GiB Does not fit even on 8 cards. Doesn't fit
Kimi-K2-Thinking 1.03T (31.7B active) Q2_K 368.3 GiB Does not fit even on 8 cards. Doesn't fit
MiMo-V2.6-Pro-MOPD 1.02T (41.0B active) Q2_K 411.6 GiB Does not fit even on 8 cards. Doesn't fit
MiMo-V2.6-Pro-RL 1.02T (41.0B active) Q2_K 411.6 GiB Does not fit even on 8 cards. Doesn't fit
Beam-1.0-501B-A23B 501B (22.7B active) Q8_0 523.6 GiB Does not fit even on 8 cards. Doesn't fit
GLM-5.2 744B (40.8B active) Q8_0 776.4 GiB Does not fit even on 8 cards. Doesn't fit
GLM-5.3 744B (40.8B active) Q8_0 776.4 GiB Does not fit even on 8 cards. Doesn't fit
Qwen3.8-2.4T-A95B 2.42T (93.3B active) Q8_0 2,524.1 GiB Does not fit even on 8 cards. Doesn't fit

The speed ceiling is memory bandwidth divided by the bytes read per generated token. It is a theoretical upper bound; public llama.cpp benchmarks typically reach about half to four-fifths of it, and small models and mixture-of-experts models often less. Public benchmarks

Detailed pages for this GPU

Other devices with 12 GB