Hardware / NVIDIA /GeForce RTX 5070 Ti

NVIDIA · GPU

GeForce RTX 5070 Ti

Runs 86 models fully in memory, 20 more with CPU offload.

Memory

16 GB

Bandwidth

896 GB/s

Runs

86

01

Specifications

Memory

16 GB

Bandwidth

896 GB/s

FP16 compute

43.9 TFLOPS

TDP

300 W

MSRP

$749

Street price

—

Released

Feb 20, 2025

Architecture

Blackwell

02

What it runs

Best quant per model — sorted by fit score.

ModelParamsBest quantContextSpeedVerdict

Ministral 3 3B 25123.8BQ8_066K178 tok/sestFits in memory Granite 4.0 Micro3.2BQ4_K_M66K314 tok/sestFits in memory UI-TARS 7B8.3BQ5_K_M66K121 tok/sestFits in memory Llama 3.2 3B Instruct3.2BQ4_K_M131K370 tok/sestFits in memory Gemma 3 4B4.3BQ4_K_M131K275 tok/sestFits in memory Llama 3.1 8B Instruct8BQ4_K_M131K148 tok/sestFits in memory Niagara 19m Batch.en20MQ4_K_M~262K59118 tok/sestFits in memory Niagara 38m Batch.en38MQ4_K_M~262K31115 tok/sestFits in memory ARK ASR 0.6B1.3BQ4_K_M~262K910 tok/sestFits in memory ARK ASR 3B4.1BQ4_K_M~262K288 tok/sestFits in memory Audio8 ASR 0.1B0.3BQ8_0~262K2249 tok/sestFits in memory Higgs Audio v3 STT2.7BQ8_0~66K250 tok/sestFits in memory Cohere Transcribe 03 20262.1BQ5_K_M~262K480 tok/sestFits in memory Distil Large v3.50.8BQ5_K_M~262K1260 tok/sestFits in memory Lite Whisper Large v3 Acc1.4BQ8_0~262K482 tok/sestFits in memory Owsm CTC v3.1 1B1.1BQ4_K_M~262K1075 tok/sestFits in memory Owsm CTC v3.2 ft 1B1BQ5_K_M~262K1008 tok/sestFits in memory Owsm CTC v4 1B1BQ4_K_M~262K1182 tok/sestFits in memory Hubert Large Ls960 ft0.3BQ8_0~262K2108 tok/sestFits in memory MMS 1b All1BQ8_0~262K675 tok/sestFits in memory Wav2vec2 Large 960h Lv60 Self0.3BQ5_K_M~262K3149 tok/sestFits in memory Hojo ASR V15.2BQ4_K_M~131K228 tok/sestFits in memory Granite 4.0 1b Speech2.3BQ4_K_M~131K514 tok/sestFits in memory Granite Speech 3.3 2b3BQ4_K_M~131K394 tok/sestFits in memory Granite Speech 4.1 2b2.3BQ8_0~131K293 tok/sestFits in memory Granite Speech 4.1 2b NAR2.3BQ4_K_M~131K514 tok/sestFits in memory STT 2.6b en2.6BQ8_0~262K260 tok/sestFits in memory Phi 4 Multimodal Instruct5.6BQ4_K_M~66K211 tok/sestFits in memory VibeVoice ASR HF8.3BQ4_K_M~131K143 tok/sestFits in memory Voxtral Mini 3B 25075BQ4_K_M~131K237 tok/sestFits in memory Voxtral Mini 4B Realtime 26024.4BQ5_K_M~66K229 tok/sestFits in memory Canary 180m Flash0.2BQ4_K_M~262K6569 tok/sestFits in memory Canary 1b1BQ4_K_M~262K1182 tok/sestFits in memory Canary 1b Flash1BQ4_K_M~262K1182 tok/sestFits in memory Canary 1b v21BQ8_0~262K675 tok/sestFits in memory Canary Qwen 2.5b2.6BQ4_K_M~262K455 tok/sestFits in memory Nemotron 3.5 ASR Streaming 0.6b0.6BQ4_K_M~262K1847 tok/sestFits in memory Nemotron Speech Streaming en 0.6b0.6BQ8_0~262K1124 tok/sestFits in memory Parakeet CTC 0.6b0.6BQ5_K_M~262K1680 tok/sestFits in memory Parakeet CTC 1.1b1.1BQ5_K_M~262K916 tok/sestFits in memory Parakeet RNNT 0.6b0.6BQ8_0~262K1124 tok/sestFits in memory Parakeet RNNT 1.1b1.1BQ4_K_M~262K1075 tok/sestFits in memory Parakeet TDT CTC 110m0.1BQ8_0~262K6133 tok/sestFits in memory Parakeet TDT 0.6b v20.6BQ8_0~262K1124 tok/sestFits in memory Parakeet TDT 0.6b v30.6BQ4_K_M~262K1971 tok/sestFits in memory Parakeet TDT 1.1b1.1BQ5_K_M~262K916 tok/sestFits in memory Nyra Health CrisperWhisper1.6BQ4_K_M~262K739 tok/sestFits in memory Whisper Large v31.5BQ5_K_M~262K672 tok/sestFits in memory Whisper Large v3 Turbo0.8BQ8_0~262K843 tok/sestFits in memory MOSS Transcribe Diarize0.9BQ4_K_M~66K1314 tok/sestFits in memory Qwen3 ASR 0.6B0.8BQ5_K_M~262K1292 tok/sestFits in memory Qwen3 ASR 0.6B HF0.8BQ5_K_M~66K1260 tok/sestFits in memory Qwen3 ASR 1.7B2BQ8_0~262K331 tok/sestFits in memory Qwen3 ASR 1.7B HF2BQ8_0~262K331 tok/sestFits in memory Zipformer cr CTC Transducer XL 290M0.3BQ5_K_M~262K3475 tok/sestFits in memory Zipformer Transducer XL 290M0.3BQ4_K_M~262K4077 tok/sestFits in memory ASR Conformer Largescaleasr0.5BQ8_0~262K1406 tok/sestFits in memory Moonshine Streaming Medium0.2BQ4_K_M~262K4927 tok/sestFits in memory Moonshine Streaming Small0.1BQ8_0~262K6746 tok/sestFits in memory Moonshine Streaming Tiny30MQ5_K_M~262K33593 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