RTX 5070 vs RTX 5090
Best sourced output tok/s per model (single-stream generation) — engine and version, quantization, context and a source link on every row. Hover a value for the quant.
RTX 50705 sourced runs · best 185.8 tok/s
RTX 509027 verified runs · best 446.7 tok/s
Specs and evidence
| RTX 5070 | RTX 5090 | |
|---|---|---|
| VRAM | 12 GB | 32 GB |
| Sourced runs on the board | 5 | 27 |
| Models measured | 5 | 17 |
| Best output tok/s | 185.8 tok/s | 446.7 tok/s |
| Highest measured peak VRAM | 9.0 GB | 31.3 GB |
| Street price (latest snapshot) | — | — |
Head-to-head
Only one model has been measured on both cards (Qwen3.8-27B). On that single data point RTX 5070 records 48.3 tok/s and RTX 5090 records 435.1 tok/s — treat it as one data point, not a ranking.
The measured matrix
| Model | RTX 5070 | RTX 5090 |
|---|---|---|
| NVIDIA-Nemotron-3.5-Lightning-30B-A3B | — | 446.7 |
| Qwen3.8-27B | 48.3 | 435.1 |
| Ornith-1.0-35B-PrismaAURA-4.75bit-vllm-MTP | — | 283.1 |
| Kimi-Linear-48B-A3B-Instruct | — | 280.8 |
| Ornith-1.5-35B-A3B | — | 258.8 |
| North-Mini-Code-1.0 | — | 258.4 |
| Recursive Flint 40B | — | 224.8 |
| Qwen3.6-35B-A3B-MTP | — | 222.9 |
| Laguna-XS-2.1 | — | 216.7 |
| KAT-Coder-V2.5-Dev-AWQ-W4A16-ASYM | — | 201.2 |
| Ornith-1.0-35B | — | 196.6 |
| NVIDIA-Nemotron-3-Nano-4B | 185.8 | — |
| Apodex-1.1-mini | 182.3 | — |
| Muse-Glimmer-30B | — | 124.8 |
| Qwen3.6-27B | — | 120.9 |
| Huihui-Qwen3.8-27B-abliterated | — | 117.7 |
| Ornith-1.5-9B | 102.0 | — |
| Qwen3.8-27B-5090-goldilocks | — | 93.5 |
| Qwen3.8-Flash-Next | 81.8 | — |
| Qwen3.6-27B-MTP | — | 28.3 |
| DeepSeek-V4-Flash-0731 | — | 26.4 |