Hardware / NVIDIA /GeForce RTX 4070 SUPER

NVIDIA · GPU

GeForce RTX 4070 SUPER

Runs 83 models fully in memory, 5 more with CPU offload. 1 of those sit close enough to 12 GB that our size estimate is what decides them.

Memory

12 GB

Bandwidth

504 GB/s

Runs

83

01

Specifications

Memory

12 GB

Bandwidth

504 GB/s

FP16 compute

35.5 TFLOPS

TDP

220 W

MSRP

$599

Street price

—

Released

Jan 17, 2024

Architecture

Ada Lovelace

02

What it runs

Best quant per model — sorted by fit score.

ModelParamsBest quantContextSpeedVerdict

Ministral 3 3B 25123.8BQ4_K_M66K175 tok/sestFits in memory Granite 4.0 Micro3.2BQ5_K_M66K151 tok/sestFits in memory Llama 3.2 3B Instruct3.2BQ4_K_M131K208 tok/sestFits in memory Gemma 3 4B4.3BQ4_K_M131K155 tok/sestFits in memory Niagara 19m Batch.en20MQ4_K_M~262K33254 tok/sestFits in memory Niagara 38m Batch.en38MQ4_K_M~262K17502 tok/sestFits in memory ARK ASR 0.6B1.3BQ4_K_M~262K512 tok/sestFits in memory ARK ASR 3B4.1BQ4_K_M~131K162 tok/sestFits in memory Audio8 ASR 0.1B0.3BQ4_K_M~262K2217 tok/sestFits in memory Higgs Audio v3 STT2.7BQ5_K_M~66K210 tok/sestFits in memory Cohere Transcribe 03 20262.1BQ5_K_M~262K270 tok/sestFits in memory Distil Large v3.50.8BQ5_K_M~262K709 tok/sestFits in memory Lite Whisper Large v3 Acc1.4BQ5_K_M~262K405 tok/sestFits in memory Owsm CTC v3.1 1B1.1BQ5_K_M~262K515 tok/sestFits in memory Owsm CTC v3.2 ft 1B1BQ8_0~262K380 tok/sestFits in memory Owsm CTC v4 1B1BQ4_K_M~262K665 tok/sestFits in memory Hubert Large Ls960 ft0.3BQ5_K_M~262K1772 tok/sestFits in memory MMS 1b All1BQ5_K_M~262K567 tok/sestFits in memory Wav2vec2 Large 960h Lv60 Self0.3BQ8_0~262K1186 tok/sestFits in memory Hojo ASR V15.2BQ4_K_M~131K128 tok/sestFits in memory Granite 4.0 1b Speech2.3BQ8_0~66K165 tok/sestFits in memory Granite Speech 3.3 2b3BQ5_K_M~66K189 tok/sestFits in memory Granite Speech 4.1 2b2.3BQ5_K_M~66K247 tok/sestFits in memory Granite Speech 4.1 2b NAR2.3BQ4_K_M~66K289 tok/sestFits in memory STT 2.6b en2.6BQ8_0~131K146 tok/sestFits in memory Voxtral Mini 3B 25075BQ4_K_M~131K133 tok/sestFits in memory Voxtral Mini 4B Realtime 26024.4BQ4_K_M~66K151 tok/sestFits in memory Canary 180m Flash0.2BQ4_K_M~262K3695 tok/sestFits in memory Canary 1b1BQ4_K_M~262K665 tok/sestFits in memory Canary 1b Flash1BQ4_K_M~262K665 tok/sestFits in memory Canary 1b v21BQ8_0~262K380 tok/sestFits in memory Canary Qwen 2.5b2.6BQ5_K_M~262K218 tok/sestFits in memory Nemotron 3.5 ASR Streaming 0.6b0.6BQ5_K_M~262K886 tok/sestFits in memory Nemotron Speech Streaming en 0.6b0.6BQ8_0~262K633 tok/sestFits in memory Parakeet CTC 0.6b0.6BQ5_K_M~262K945 tok/sestFits in memory Parakeet CTC 1.1b1.1BQ5_K_M~262K515 tok/sestFits in memory Parakeet RNNT 0.6b0.6BQ8_0~262K633 tok/sestFits in memory Parakeet RNNT 1.1b1.1BQ4_K_M~262K605 tok/sestFits in memory Parakeet TDT CTC 110m0.1BQ8_0~262K3450 tok/sestFits in memory Parakeet TDT 0.6b v20.6BQ8_0~262K633 tok/sestFits in memory Parakeet TDT 0.6b v30.6BQ4_K_M~262K1109 tok/sestFits in memory Parakeet TDT 1.1b1.1BQ5_K_M~262K515 tok/sestFits in memory Nyra Health CrisperWhisper1.6BQ4_K_M~262K416 tok/sestFits in memory Whisper Large v31.5BQ5_K_M~262K378 tok/sestFits in memory Whisper Large v3 Turbo0.8BQ8_0~262K474 tok/sestFits in memory MOSS Transcribe Diarize0.9BQ4_K_M~66K739 tok/sestFits in memory Qwen3 ASR 0.6B0.8BQ5_K_M~262K727 tok/sestFits in memory Qwen3 ASR 0.6B HF0.8BQ8_0~66K474 tok/sestFits in memory Qwen3 ASR 1.7B2BQ8_0~262K186 tok/sestFits in memory Qwen3 ASR 1.7B HF2BQ8_0~262K186 tok/sestFits in memory Zipformer cr CTC Transducer XL 290M0.3BQ8_0~262K1309 tok/sestFits in memory Zipformer Transducer XL 290M0.3BQ5_K_M~262K1955 tok/sestFits in memory ASR Conformer Largescaleasr0.5BQ8_0~262K791 tok/sestFits in memory Moonshine Streaming Medium0.2BQ8_0~262K1581 tok/sestFits in memory Moonshine Streaming Small0.1BQ5_K_M~262K5669 tok/sestFits in memory Moonshine Streaming Tiny30MQ5_K_M~262K18896 tok/sestFits in memory Moonshine Tiny30MQ8_0~262K12649 tok/sestFits in memory GLM ASR Nano 25122.3BQ5_K_M~131K247 tok/sestFits in memory Llama 3.2 1B Instruct1.2BQ8_033K316 tok/sestFits in memory Phi 4 Multimodal Instruct5.6BQ4_K_M~33K119 tok/sestFits in memory

A speed marked est is calculated from this card's memory bandwidth and the size of the weights it has to read per word. It is not a measurement.

Something wrong on this page? Tell us