Live rates — RunPod Secure vs RunPod Community vs Vast.ai Verified
| GPU | VRAM (GB) | RunPod Secure ($/hr) | RunPod Community ($/hr) | Vast.ai Verified ($/hr) | Lowest ($/hr) | Vast live rigs | RunPod Secure 730h | Vast 730h | Deploy |
|---|---|---|---|---|---|---|---|---|---|
| RTX 4090 24GB | 24 | $0.89 | $0.34 | $0.46 | $0.34 | 3 | $649.70 | $333.10 | RunPodVast.ai |
| RTX 3090 24GB | 24 | $0.50 | $0.22 | $0.27 | $0.22 | 2 | $365.00 | $197.03 | RunPodVast.ai |
| RTX A5000 24GB | 24 | $0.27 | $0.16 | $0.23 | $0.16 | 1 | $197.10 | $170.02 | RunPodVast.ai |
Dual/quad reference for RTX 4090 24GB: RunPod 2x from $0.68/hr, Vast.ai 2x from $0.91/hr, lowest 2x floor $0.68/hr. 4x floors: RunPod $1.36, Vast.ai $1.83, lowest $1.36/hr.
Open-model compatibility — 1x vs 2x at 32K context
| Open model (32K ctx) | RTX 4090 24GB 1x | RTX 4090 24GB 2x | RTX 3090 24GB 1x | RTX 3090 24GB 2x | RTX A5000 24GB 1x | RTX A5000 24GB 2x |
|---|---|---|---|---|---|---|
| Qwen3.8-27B | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) |
| Qwen3.8-Flash-Next | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| Arete-27B | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) |
| diffusiongemma-26B-A4B-it | Requires 2x+ Requires 2x+ Fits INT4 (32K) | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) | Requires 2x+ Requires 2x+ Fits INT4 (32K) | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) | Requires 2x+ Requires 2x+ Fits INT4 (32K) | No fit at 2x Fits FP8 (32K) Fits INT4 (32K) |
| MiniMax-M3.1-Flash | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| GLM-5.3-Flash | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| GLM-5.3 | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| GLM-5.2 | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| DeepSeek-V4.1-Flash-1017 | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| DeepSeek-V4.1-Flash | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| DeepSeek-V4-Flash-0731 | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| DeepSeek-V4-Pro-0813 | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| Kimi-K2.7-Code | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| Kimi-K3.1-Pro | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| Qwen3.8-2.4T-A95B | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
| DeepSeek-V4-Flash-DSpark | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x | Requires 2x+ Requires 2x+ Requires 2x+ | No fit at 2x No fit at 2x No fit at 2x |
Verdicts hard-computed from each repo's safetensors and config: FP16 (16-bit), FP8 (8-bit) and INT4 (4.5-bit effective) weights + GQA KV cache at 32768 tokens + runtime overhead, concurrency 1.
What the monthly gap looks like
Rent a RTX 4090 24GB around the clock (730 h/month): RunPod Secure bills $649.70/mo, while the cheapest verified, rentable Vast.ai bundle bills $333.10/mo — a real gap of $316.60 every month ($3799.21 per year) for the identical GPU class. Locking a yearly hyperscaler contract instead of checking this table is where the real money leaks.