Skip to content

What can the RTX Spark run?

What fits on the RTX Spark
Text models: 119Visual models: 5Strongest first
Memory
128 GB unified
Bandwidth
300 GB/s
Maker
NVIDIA
Also called
N1X

Open the calculator on this machine

Text models that run on it

Text models that run on the RTX Spark, best first
ModelBest fitContextSpeedStrongest at
GLM-5.3 Flash2-bit104K28 tok/sChat, Coding, Agents, Docs
Qwen3.8 Flash-Next4-bit256K48 tok/sChat, Coding, Agents
MiMo V2.6 Flash2-bit256K34 tok/sCoding, Agents
MiniMax M31.6-bit256K38 tok/sDocs
DeepSeek V4 Flash2-bit1M28 tok/sDocs
Hy32-bit32K24 tok/sChat, Coding, Agents, Docs
Qwen3.8 27B8-bit256K7.3 tok/sChat, Coding, Agents, Docs
Nemotron 3 Ultra 550B A55B1.6-bit256K16 tok/sChat, Coding, Docs
Qwen3.5 397B A17B1.6-bit256K47 tok/sAgents, Docs
MiniMax M2.73-bit36K37 tok/sCoding, Agents
Muse Glimmer 30B8-bit128K6.9 tok/sDocs
Mistral Medium 3.5 128B6-bit44K2.0 tok/sChat, Coding, Agents, Docs
Qwen3.5 122B A10B6-bit256K23 tok/sChat, Coding, Agents, Docs
Nemotron 3 Super 120B A12B6-bit256K19 tok/sChat, Coding, Agents, Docs
Nemotron 3.5 Bolt 74B A9B8-bit256K19 tok/sChat, Coding, Agents, Docs
Qwen3.6 35B A3B8-bit256K56 tok/sChat, Coding, Agents, Docs
Gemma 4 31B8-bit256K6.2 tok/sChat, Coding, Agents, Docs
GLM-4.7 Flash8-bit198K56 tok/sChat, Coding, Agents
Mistral Small 4 119BFP848K30 tok/sChat, Docs
gpt-oss 120BMXFP4128K51 tok/sChat, Coding
Qwen3 Coder Next 80B8-bit256K54 tok/sCoding, Agents
Nemotron 3.5 Lightning 30B A3B8-bit256K54 tok/sAgents, Docs
Llama 3.3 70B8-bit128K2.8 tok/sChat, Coding, Agents, Docs
Olmo 3.1 32B8-bit64K6.2 tok/sChat, Coding, Agents, Docs
Qwen3 32B8-bit40K6.1 tok/sNot rated
Qwen3.6 27B8-bit256K7.3 tok/sText and images
Qwen3 Coder 30B A3B Instruct8-bit256K55 tok/sCoding
Qwen3.5 27B8-bit256K7.3 tok/sText and images
Qwen3.5 35B A3B8-bit256K56 tok/sText and images
NVIDIA Nemotron 3 Nano 30B A3B BF168-bit256K54 tok/sNot rated
GLM 4.5 Air8-bit20K13 tok/sNot rated
Inkling Small2-bit4K44 tok/sText and images
Qwen3 30B A3B8-bit40K55 tok/sNot rated
MiniMax M2.53-bit36K33 tok/sNot rated
Qwen3 30B A3B Instruct 25078-bit256K55 tok/sNot rated
Laguna XS.28-bit256K57 tok/sNot rated
OTel 2.0 LLM 31B IT8-bit256K6.2 tok/sNot rated
Laguna M.13-bit36K17 tok/sNot rated
Qwen2.5 32B Instruct8-bit32K6.1 tok/sNot rated
MiniCPM V 4.7 35B A3B8-bit256K56 tok/sText and images
Laguna XS 2.18-bit256K57 tok/sNot rated
Qwen2.5 VL 32B Instruct8-bit125K6.0 tok/sText and images
Qwen2.5 Coder 32B Instruct8-bit32K6.1 tok/sCoding
Laguna S 2.16-bit480K28 tok/sNot rated
Qwen3 VL 30B A3B Instruct8-bit256K55 tok/sText and images
DeepSeek R1 Distill Qwen 32B8-bit128K6.1 tok/sNot rated
Gemma 3 27B IT8-bit128K7.5 tok/sText and images
Qwen3 VL 235B A22B Instruct3-bit32K18 tok/sText and images
Qwen3 235B A22B3-bit32K18 tok/sNot rated
Ornith 1.5 35B A3B8-bit256K56 tok/sNot rated
Ornith 1.5 122B6-bit256K19 tok/sNot rated
Ornith 1.5 397B1.6-bit256K35 tok/sNot rated
Qwen3 VL 32B Instruct8-bit256K6.0 tok/sText and images
Step 3.5 Flash3-bit56K41 tok/sNot rated
Qwen2.5 72B Instruct8-bit32K2.7 tok/sNot rated
Qwen3 Next 80B A3B Instruct8-bit256K54 tok/sNot rated
Llama 4 Scout 17B 16E Instruct8-bit28K11 tok/sText and images
MiniMax M23-bit36K33 tok/sNot rated
MiMo V2.52-bit224K29 tok/sNot rated
Qwen2.5 VL 72B Instruct8-bit125K2.7 tok/sText and images

Best fit is the highest quality, up to 8-bit, that leaves the use case’s context. Ranked by our use-case scores, then by Hugging Face downloads. Speeds are rough, plus or minus 25%. The top 60 of 119 are shown.

Image and video models

Image and video models that run on the RTX Spark
ModelBest fitPer outputStrongest at
MiniMax H316-bit16 minVideo
LTX-2.516-bit1.8 minVideo
Qwen-Image16-bit9.2 minImages
FLUX.2 [dev]16-bit4.1 minImages
Qwen Image 251216-bit9.2 minImages

Only just too big

These need up to a quarter more memory than the RTX Spark has. Spilling into system RAM or a lower precision can get them running, slowly.

Questions

What is the best LLM for the RTX Spark?

Of the open models we track, the strongest that run well on the RTX Spark are GLM-5.3 Flash, Qwen3.8 Flash-Next, MiMo V2.6 Flash. 119 text models fit it in all.

Calculated from published specs, not benchmarked. All machines