RTX 5070
A discrete graphics card in the NVIDIA RTX 50 series, positioned in the upper-mid-range segment. It features 12 GB of video memory and supports AI-enhanced technologies such as DLSS.
Core Facts
In an RTX 5070 (12GB VRAM) configuration, the local Qwen3 9B model's output speed reached approximately 70–76 tokens/second.
The RTX 5070 belongs to NVIDIA's Blackwell architecture consumer graphics card product line and features approximately 12 GB of VRAM.
The RTX 5090 is based on NVIDIA's Blackwell architecture.
Timeline (last 90 days)
Under Strata 0.2.4, DeepSeek V4.1 Flash generates at about 5 tok/s on an RTX 5070, while an RTX 3060 instead reaches 8 to 10 tok/s.
SteerTTS achieves an RTFx of 60–80 on an RTX 5070, and 4–8 on CPU without the refiner.
Demon's Souls: Remake on KyTyPS5 and its forks ran at only 1–2 FPS on top-tier hardware in early September, and by mid-October a custom build ran at over 60 FPS at native 1440p on an RTX 5070 + Ryzen 7 9800X3D, approaching 120 FPS with DLSS frame generation enabled, and by the middle of the month it was completable.
The Strata model generates 94 tokens per second on an RTX 5070 and 60 tokens per second on an AMD 9070 XT.
In testing, the 3-bit quantized version outputs at 53 tokens per second.
In testing, an RTX 5070 running the 2-bit quantized version outputs 94 tokens per second.
In Strata's speed table, under a 128k context, Q2_0 (2-bit) achieves 65.1 tokens/s and IQ2_XS achieves 52 tokens/s; under a 4k short context, IQ3_XXS achieves 65.6 tokens/s.
With a 5070 12 GB GPU paired with 64 GB of memory and MTP enabled, generation speed can exceed 50 tokens/s, with official tests reaching 60–95 tokens/s.
With a 5070 12 GB GPU paired with 64 GB of memory, speed can reach 93 tokens/s in short chat scenarios and 74 tokens/s for 128K long contexts.
Strata is an open-source inference runtime capable of running the 125-billion-parameter Qwen3.8 Flash Next model on an RTX 5070 (12 GB VRAM) at an output speed of approximately 90 tokens per second.
13 more timeline events
All Facts (20)
In an RTX 5070 (12GB VRAM) configuration, the local Qwen3 9B model's output speed reached approximately 70–76 tokens/second.
70%VerifiedThe RTX 5070 belongs to NVIDIA's Blackwell architecture consumer graphics card product line and features approximately 12 GB of VRAM.
65%VerifiedThe RTX 5090 is based on NVIDIA's Blackwell architecture.
65%UnverifiedThe primary test platform for the Fast H3 workflow is a high-end environment consisting of an RTX 5090 paired with 128GB of memory.
60%UnverifiedThe UP host's test environment consisted of an RTX 5070 GPU (12GB VRAM), the Qwen3 series 9B parameter model, and a 64K context length.
60%UnverifiedUnder Strata 0.2.4, DeepSeek V4.1 Flash generates at about 5 tok/s on an RTX 5070, while an RTX 3060 instead reaches 8 to 10 tok/s.
50%UnverifiedSteerTTS achieves an RTFx of 60–80 on an RTX 5070, and 4–8 on CPU without the refiner.
50%UnverifiedThe Strata model generates 94 tokens per second on an RTX 5070 and 60 tokens per second on an AMD 9070 XT.
50%UnverifiedIn testing, the 3-bit quantized version outputs at 53 tokens per second.
50%UnverifiedIn testing, an RTX 5070 running the 2-bit quantized version outputs 94 tokens per second.
50%UnverifiedIn Strata's speed table, under a 128k context, Q2_0 (2-bit) achieves 65.1 tokens/s and IQ2_XS achieves 52 tokens/s; under a 4k short context, IQ3_XXS achieves 65.6 tokens/s.
50%UnverifiedWith a 5070 12 GB GPU paired with 64 GB of memory and MTP enabled, generation speed can exceed 50 tokens/s, with official tests reaching 60–95 tokens/s.
50%UnverifiedWith a 5070 12 GB GPU paired with 64 GB of memory, speed can reach 93 tokens/s in short chat scenarios and 74 tokens/s for 128K long contexts.
50%UnverifiedStrata is an open-source inference runtime capable of running the 125-billion-parameter Qwen3.8 Flash Next model on an RTX 5070 (12 GB VRAM) at an output speed of approximately 90 tokens per second.
50%UnverifiedOn an RTX 5070 with a 6-core Ryzen and 64 GB of memory, Strata achieves approximately 90 tokens/second for short conversations, approximately 67 tokens/second at a 128,000-token context, and approximately 1,300 tokens/second for reading prompts.
50%UnverifiedCompared to the previous-generation RTX 4070 series, the RTX 5070 shows improvements in shader core count, VRAM bandwidth, and AI compute performance
50%UnverifiedThe desktop features an RTX 5070 graphics card and 32GB of 6000MHz memory
50%UnverifiedThe Neo X1 Pro paired with the RTX 5070 leads the Minisforum G1 Pro (RTX 5060 + Ryzen 8945HX) by approximately 73% in 3DMark scores.
50%UnverifiedThe Neo X1 Pro paired with the RTX 5070 achieves an average of 93 FPS and a minimum of 51 FPS in Cyberpunk 2077 at 1440p with RTX Ultra preset and DLSS Transformer mode.
50%UnverifiedThe RTX 5050 is the entry-level mobile model in the Blackwell architecture, positioned below the RTX 5060 and RTX 5070.
50%