Your rig

RTX 6000 Ada · 48 GB

64 GB system RAM · Ada Lovelace · 960 GB/s · 46.4 GB usable VRAM
Edit hardware Export as JSON

Context length

KV cache at f16, 32K tokens

Quality floor
Q8_0 — near-lossless Q6_K Q5_K_M Q4_K_M — recommended Q3_K_M Q2_K — damaged
KV cache type
System RAM (GB)
Runtime
Ollamallama.cppLM StudiovLLM

4 of 5 runtimes drive this device.

Allow CPU offload Include partial-offload fits

Offloaded layers run at 60 GB/s host bandwidth — expect single-digit tokens per second.

Which one should I actually run?

Best models for your RTX 6000 Ada

Updated 8 Oct 2026 1 new this month

What do you want to do? Each row is the best-ranked model on that use case's shortlist that fits at Q8_0 and 32K. The ordering is editorial; the sizes and speeds are computed. How the shortlists work.

RecommendationModelWhyQuant · totalTok/s est.
Best overall Qwen3.8 27B
Alibaba · Apache 2.0
Best capability per gigabyte: a current-generation dense 27B with vision and 262K context. Q8_0 · 30.1 GB 21 Runs great
Best for coding Qwen3.8 27B
Alibaba · Apache 2.0
Terminal-Bench 73, DeepSWE 42 — a generation ahead of anything else that fits 24 GB. Q8_0 · 30.1 GB 21 Runs great
Best reasoning Gemma 4 31B
Google · Apache 2.0
89% AIME — the strongest maths of any model under 32B. Q8_0 · 37.4 GB 19 Runs great
Best for writing Gemma 4 31B
Google · Apache 2.0
The cleanest, least verbose prose of the 2026 24–32 GB class. Q8_0 · 37.4 GB 19 Runs great
Best vision Qwen3.8 27B
Alibaba · Apache 2.0
Image and video in, OSWorld 84 — the strongest local vision-language model at 24 GB. Q8_0 · 30.1 GB 21 Runs great
Best for agents Qwen3.8 27B
Alibaba · Apache 2.0
Computer use and tool calling are what 3.8 was trained for (OSWorld 84). Q8_0 · 30.1 GB 21 Runs great
Best translation Qwen3.8 27B
Alibaba · Apache 2.0
119 languages, and Qwen translations read as idiomatic rather than literal. Q8_0 · 30.1 GB 21 Runs great
Fastest good model Qwen3.6 35B-A3B
Alibaba · Apache 2.0
3.3B active per token: the fastest model that still competes with dense 27Bs. Q8_0 · 36.7 GB 68 Runs great
Best long context Qwen3.8 27B
Alibaba · Apache 2.0
262K native, and only 16 of 64 blocks keep a KV cache — long context is cheap here. Q8_0 · 36.1 GB
at 128K context
21 Runs great

Everything that fits

Runs great

54 of 80 · fits with headroom for a longer prompt
ModelParamsQuantWeights +KV 32KTotal VRAM headroom Tok/sRuntime
Qwen3.6 35B-A3B 35.9B MoE / 3.3B act Q8_0 35.5 GB 0.63 GB 36.7 GB
9.7 GB free
68 Ollama Run
Ornith 1.5 35B-A3B 35.9B MoE / 3.3B act Q8_0 35.5 GB 0.63 GB 36.7 GB
9.7 GB free
68 Ollama Run
Olmo 3.1 32B Instruct 32.2B Q8_0 31.9 GB 2.75 GB 35.2 GB
11.2 GB free
18 Ollama Run
Nemotron 3.5 Lightning 30B-A3B 31.6B MoE / 3.2B act Q8_0 31.3 GB 0.19 GB 32.1 GB
14.3 GB free
71 Ollama Run
Gemma 4 31B 31.3B Q8_0 31.0 GB 5.78 GB 37.4 GB
9.0 GB free
19 Ollama Run
GLM-4.7-Flash 30B-A3B 31.2B MoE / 3B act Q8_0 30.9 GB 1.65 GB 33.1 GB
13.3 GB free
75 Ollama Run
Qwen3 30B-A3B 30.5B MoE / 3.3B act Q8_0 30.2 GB 3.00 GB 33.8 GB
12.6 GB free
68 Ollama Run
Qwen3 Coder 30B-A3B 30.5B MoE / 3.3B act Q8_0 30.2 GB 3.00 GB 33.8 GB
12.6 GB free
68 Ollama Run
Muse Glimmer 30B 29.8B Q8_0 29.5 GB 0.48 GB 30.6 GB
15.8 GB free
20 Ollama Run
Granite 4.1 30B 28.9B Q8_0 28.6 GB 8.00 GB 37.2 GB
9.2 GB free
20 Ollama Run
Qwen3.8 27B 27.8B Q8_0 27.5 GB 2.00 GB 30.1 GB
16.3 GB free
21 Ollama Run
Qwen3.6 27B 27.8B Q8_0 27.5 GB 2.00 GB 30.1 GB
16.3 GB free
21 Ollama Run
Gemma 3 27B 27.4B Q8_0 27.1 GB 2.91 GB 30.6 GB
15.8 GB free
21 Ollama Run
Gemma 4 26B-A4B 26.5B MoE / 3.8B act Q8_0 26.2 GB 1.45 GB 28.3 GB
18.1 GB free
59 Ollama Run
Devstral Small 2 24B 24B Q8_0 23.7 GB 5.00 GB 29.3 GB
17.1 GB free
24 llama.cpp Run
Mistral Small 3.2 24B 23.6B Q8_0 23.4 GB 5.00 GB 29.0 GB
17.4 GB free
25 Ollama Run
gpt-oss 20B 20.9B MoE / 3.6B act MXFP4 10.8 GB 0.75 GB 12.2 GB
34.2 GB free
120 Ollama Run
Qwen3 14B 14.8B Q8_0 14.6 GB 5.00 GB 20.2 GB
26.2 GB free
40 Ollama Run
DeepSeek-R1-Distill-Qwen 14B 14.8B Q8_0 14.6 GB 6.00 GB 21.2 GB
25.2 GB free
40 Ollama Run
Qwen2.5-Coder 14B 14.8B Q8_0 14.6 GB 6.00 GB 21.2 GB
25.2 GB free
40 Ollama Run
Phi-4 14B 14.7B Q8_0 14.5 GB 3.13 GB 18.3 GB
28.1 GB free
40 Ollama Run
Ministral 3 14B 13.9B Q8_0 13.8 GB 5.00 GB 19.4 GB
27.0 GB free
42 Ollama Run
Gemma 3 12B 12.2B Q8_0 12.1 GB 2.31 GB 15.0 GB
31.4 GB free
48 Ollama Run
Mistral NeMo 12B 12.2B Q8_0 12.1 GB 5.00 GB 17.7 GB
28.7 GB free
48 Ollama Run
Gemma 4 12B 12B Q8_0 11.9 GB 2.31 GB 14.8 GB
31.6 GB free
49 Ollama Run
Qwen3.5 9B 9.65B Q8_0 9.5 GB 1.00 GB 11.1 GB
35.3 GB free
61 Ollama Run
Ornith 1.5 9B 9.41B Q8_0 9.3 GB 1.00 GB 10.9 GB
35.5 GB free
62 Ollama Run
Ministral 3 8B 8.92B Q8_0 8.8 GB 4.25 GB 13.7 GB
32.7 GB free
66 Ollama Run
Granite 4.1 8B 8.79B Q8_0 8.7 GB 5.00 GB 14.3 GB
32.1 GB free
67 Ollama Run
LFM2.5 8B-A1B 8.47B MoE / 1.5B act Q8_0 8.4 GB 0.38 GB 9.4 GB
37.0 GB free
151 Ollama Run
Fara 7B 8.29B Q8_0 8.2 GB 1.75 GB 10.6 GB
35.8 GB free
71 llama.cpp Run
Qwen3 8B 8.19B Q8_0 8.1 GB 4.50 GB 13.2 GB
33.2 GB free
72 Ollama Run
Llama 3.1 8B Instruct 8.03B Q8_0 7.9 GB 4.00 GB 12.5 GB
33.9 GB free
73 Ollama Run
Gemma 4 E4B 8.0B Q8_0 7.9 GB 0.47 GB 9.0 GB
37.4 GB free
73 Ollama Run
Ling 3.0 Tiny 7.9B-A1.3B 7.9B MoE / 1.3B act Q8_0 7.8 GB 0.21 GB 8.6 GB
37.8 GB free
174 llama.cpp Run
DeepSeek-R1-Distill-Qwen 7B 7.62B Q8_0 7.5 GB 1.75 GB 9.9 GB
36.5 GB free
77 Ollama Run
Qwen2.5-Coder 7B 7.62B Q8_0 7.5 GB 1.75 GB 9.9 GB
36.5 GB free
77 Ollama Run
Olmo 3 7B Instruct 7.3B Q8_0 7.2 GB 5.50 GB 13.3 GB
33.1 GB free
80 Ollama Run
Mistral 7B Instruct v0.3 7.25B Q8_0 7.2 GB 4.00 GB 11.8 GB
34.6 GB free
81 Ollama Run
Gemma 4 E2B 5.1B Q8_0 5.0 GB 0.23 GB 5.9 GB
40.5 GB free
115 Ollama Run
Qwen3.5 4B 4.66B Q8_0 4.6 GB 1.00 GB 6.2 GB
40.2 GB free
126 Ollama Run
Gemma 3 4B 4.3B Q8_0 4.3 GB 0.86 GB 5.7 GB
40.7 GB free
137 Ollama Run
Qwen3 4B 4.02B Q8_0 4.0 GB 4.50 GB 9.1 GB
37.3 GB free
146 Ollama Run
Ministral 3 3B 3.85B Q8_0 3.8 GB 3.25 GB 7.7 GB
38.7 GB free
153 Ollama Run
Phi-4-mini 3.8B 3.84B Q8_0 3.8 GB 4.00 GB 8.4 GB
38.0 GB free
153 Ollama Run
Granite 4.1 3B 3.4B Q8_0 3.4 GB 2.50 GB 6.5 GB
39.9 GB free
173 Ollama Run
Llama 3.2 3B Instruct 3.21B Q8_0 3.2 GB 3.50 GB 7.3 GB
39.1 GB free
183 Ollama Run
LFM2.5 2.6B 2.7B Q8_0 2.7 GB 0.50 GB 3.8 GB
42.6 GB free
218 llama.cpp Run
Qwen3.5 2B 2.27B Q8_0 2.2 GB 0.38 GB 3.2 GB
43.2 GB free
259 Ollama Run
Qwen3 1.7B 1.72B Q8_0 1.7 GB 3.50 GB 5.8 GB
40.6 GB free
341 Ollama Run
Llama 3.2 1B Instruct 1.24B Q8_0 1.2 GB 1.00 GB 2.8 GB
43.6 GB free
474 Ollama Run
Gemma 3 1B 1.0B Q8_0 1.0 GB 0.14 GB 1.7 GB
44.7 GB free
587 Ollama Run
Qwen3.5 0.8B 0.87B Q8_0 0.9 GB 0.38 GB 1.8 GB
44.6 GB free
675 Ollama Run
Qwen3 0.6B 0.6B Q8_0 0.6 GB 3.50 GB 4.7 GB
41.7 GB free
979 Ollama Run

Runs — tight

4 models · fits at 32K, drop context or quant to go longer
ModelParamsQuantWeights +KV 32KTotal VRAM headroom Tok/sRuntime
LLM-jp 4 33B Thinking 33.2B Q8_0 32.9 GB 8.00 GB 41.5 GB
4.9 GB free
18 llama.cpp Run
Qwen3 32B 32.8B Q8_0 32.5 GB 8.00 GB 41.1 GB
5.3 GB free
18 llama.cpp Run
DeepSeek-R1-Distill-Qwen 32B 32.8B Q8_0 32.5 GB 8.00 GB 41.1 GB
5.3 GB free
18 llama.cpp Run
Qwen2.5-Coder 32B 32.8B Q8_0 32.5 GB 8.00 GB 41.1 GB
5.3 GB free
18 llama.cpp Run

Won't fit in VRAM

22 models · some are possible with CPU offload
ModelParamsQuantWeights +KV 32KTotal Over budget Tok/sRuntime
gpt-oss 120B 117B MoE / 5.1B act MXFP4 60.5 GB 1.13 GB 62.2 GB +15.8 GB ~17 llama.cpp Why
Llama 3.3 70B Instruct 70.6B Q8_0 69.9 GB 10.00 GB 80.5 GB +34.1 GB ~1.0 llama.cpp Why
DeepSeek-R1-Distill-Llama 70B 70.6B Q8_0 69.9 GB 10.00 GB 80.5 GB +34.1 GB ~1.0 llama.cpp Why
Kimi K3 2.8T-A104B 2780B MoE / 104B act Q8_0 2750.9 GB 0.84 GB 2752.3 GB +2705.9 GB ~0.1 llama.cpp Why
Qwen3.8 2.4T-A95B 2446B MoE / 95B act Q8_0 2420.4 GB 2.88 GB 2423.9 GB +2377.5 GB ~0.2 llama.cpp Why
DeepSeek V4 Pro 1.6T-A49B 1650B MoE / 49B act Q8_0 1632.7 GB 2.14 GB 1635.5 GB +1589.1 GB ~0.3 llama.cpp Why
Kimi K2.6 1T-A32B 1027B MoE / 32B act Q8_0 1016.2 GB 2.14 GB 1019.0 GB +972.6 GB ~0.5 llama.cpp Why
GLM-5.2 744B-A40B 753B MoE / 40B act Q8_0 745.1 GB 2.74 GB 748.5 GB +702.1 GB ~0.4 llama.cpp Why
GLM-5.3 744B-A40B 753B MoE / 40B act Q8_0 745.1 GB 2.74 GB 748.5 GB +702.1 GB ~0.4 llama.cpp Why
DeepSeek-R1 671B 671B MoE / 37B act Q8_0 664.0 GB 2.14 GB 666.7 GB +620.3 GB ~0.4 llama.cpp Why
DeepSeek V4.1 Flash 552B-A16B New 552B MoE / 16B act Q8_0 546.2 GB 0.15 GB 547.0 GB +500.6 GB ~1.0 vLLM Why
Qwen3.5 397B-A17B 403B MoE / 17B act Q8_0 398.8 GB 0.94 GB 400.3 GB +353.9 GB ~0.9 llama.cpp Why
Ornith 1.5 397B-A17B 397B MoE / 17B act Q8_0 392.8 GB 0.94 GB 394.4 GB +348.0 GB ~0.9 llama.cpp Why
GLM-5.3-Flash 320B-A18B 321B MoE / 18B act Q8_0 317.6 GB 0.34 GB 318.6 GB +272.2 GB ~0.9 llama.cpp Why
DeepSeek V4 Flash 284B-A13B 284B MoE / 13B act Q8_0 281.0 GB 1.51 GB 283.1 GB +236.7 GB ~1.3 llama.cpp Why
Qwen3 235B-A22B 235B MoE / 22B act Q8_0 232.5 GB 5.88 GB 239.0 GB +192.6 GB ~0.8 llama.cpp Why
Qwen3.8-Flash-Next 180B-A6B 180B MoE / 6B act Q8_0 178.1 GB 0.75 GB 179.5 GB +133.1 GB ~3.1 llama.cpp Why
Qwen3.5 122B-A10B 125B MoE / 10B act Q8_0 123.7 GB 0.75 GB 125.0 GB +78.6 GB ~2.1 llama.cpp Why
Nemotron 3 Super 120B-A12B 124B MoE / 12B act Q8_0 122.7 GB 0.25 GB 123.6 GB +77.2 GB ~1.8 llama.cpp Why
Ling 3.0 Flash 124B-A5B 124B MoE / 5.1B act Q8_0 122.7 GB 0.25 GB 123.5 GB +77.1 GB ~4.2 llama.cpp Why
Mistral Small 4 119B-A6B 119B MoE / 6.5B act Q8_0 117.8 GB 0.70 GB 119.1 GB +72.7 GB ~3.4 llama.cpp Why
Llama 4 Scout 109B-A17B 109B MoE / 17B act Q8_0 107.9 GB 2.63 GB 111.1 GB +64.7 GB ~1.3 llama.cpp Why

Totals = quantised weights + f16 KV cache at 32K tokens + 0.6 GB runtime overhead, against 46.4 GB usable VRAM (48 GB less display and driver reserve). Tokens per second are estimated from 960 GB/s peak memory bandwidth at batch 1 — not measured. Read the method.


Same memory, different speed

These devices have the same 46.4 GB usable, so they run exactly the models above. Generation speed scales with memory bandwidth.

DeviceBandwidthSpeed vs yours
L40S
Ada Lovelace · 2023
864 GB/s 0.90×
Radeon Pro W7900
RDNA 3 · 2023
864 GB/s 0.90×
RTX A6000
Ampere · 2020
768 GB/s 0.80×

Best LLM for 48 GB VRAM, with every card in the class →