RTX 5070

NVIDIA · 12 GB · 672 GB/s

6 models7 model configurations9 exact recipes

Representative image: 5070-family og:image depicts the 5070 Ti FE; representative for the 5070 (same Blackwell FE chassis family). — 5070-family og:image depicts the 5070 Ti FE; representative for the 5070 (same Blackwell FE chassis family). NVIDIA ›

Model configurations

Each card is a model and hardware setup, including any documented host RAM, storage or CPU requirements. Select a recipe to update its speeds, context and exact command.

Search configurations

Sort by

7 configurations · 9 recipes

Gemma 4 12B

RTX 507012 GB1 card per machine · 1 machine

TabbyAPIEXL3 4 bpw

58.7decode tok/s

1,452prefill tok/s

64Kcontext

Tested on this card · Sep 27, 2026

Full recipe ›

Gemma 4 26B A4B

RTX 507012 GB1 card per machine · 1 machine

Host RAM: ≥ 14.6 GB free
Disk: ≥ 18.67 GB NVMe

llama.cppFormat not recorded

35decode tok/s

1,460prefill tok/s

128Kcontext

Tested on this card · Oct 8, 2026

Full recipe ›

Ornith-1.5-35B-A3B

RTX 507012 GB1 card per machine · 1 machine

Host RAM: ≥ 21.7 GB free
Disk: ≥ 22.62 GB NVMe

llama.cppFormat not recorded

32.3decode tok/s

818prefill tok/s

256Kcontext

Tested on this card · Oct 8, 2026

Full recipe ›

Qwen3.6-35B-A3B

RTX 507012 GB1 card per machine · 1 machine

Host RAM: ≥ 20.2 GB free
Disk: ≥ 22.85 GB NVMe

llama.cppFormat not recorded

39.4decode tok/s

876prefill tok/s

256Kcontext

Tested on this card · Oct 8, 2026

Full recipe ›

Qwen3.8-27B

RTX 507012 GB1 card per machine · 1 machine

Exact recipe

59decode tok/s

1,009prefill tok/s

32Kcontext

Tested on this card · Oct 10, 2026

Full recipe ›

Qwen3.8-Flash-Next

RTX 507012 GB1 card per machine · 1 machine

Host RAM: ≥ 18.6 GB free
Disk: ≥ 80 GB NVMe

strataFormat not recorded

30.1decode tok/s

807prefill tok/s

64Kcontext

Tested on this card · Oct 8, 2026

Full recipe ›

Qwen3.8-Flash-Next

RTX 507012 GB1 card per machine · 1 machine

Host RAM: ≥ 48 GB free
Disk: ≥ 64 GB NVMe

SGLangEXL3 2.05 bpw

56decode tok/s

1,324prefill tok/s

128Kcontext

Tested on this card · Oct 16, 2026

Full recipe ›

Speeds are the selected recipe’s recorded measurements, not estimates; “—” means unmeasured. Evidence distinguishes checks on this card, sibling-card checks, earlier checks and publisher reports.

Hardware

Memory	12 GB GDDR7
Bandwidth	672 GB/s
Architecture	NVIDIA Blackwell
CUDA cores	6,144
AI TOPS	988

Precision	Dense TFLOPS	2:4 sparse
FP4	493.9	987.8
FP8	246.9	493.9
INT8	246.9	493.9
FP16	123.5	246.9
BF16	61.7	123.5
TF32	30.9	61.7
FP32	30.9	–

MSRP	$549
Street price	$799.99
Availability	In stock

Listing	Price	Seen	Where
NVIDIA GeForce RTX 5070 Overclocked Triple Fan 12GB GDDR7 PCIe 5.0 Graphics Card	$799.99	2026-09-22	microcenter ›
MSI SHADOW GeForce RTX 5070 12GB GDDR7 SHADOW 3X OC	$609.99	2026-08-31	newegg ›
MSRP	$549	2025-02	NVIDIA ›

Sources: NVIDIA, NVIDIA (2), newegg, microcenter