AMD Instinct MI350X 288GB, what it runs and what it costs to rent
Specs checked · Rental prices from · Aliteq Team · How we compute this
Some links on this page are referral links: we may earn a commission, at no cost to you.
The short answer
The AMD Instinct MI350X has 288 GB of VRAM and 8000 GB/s of memory bandwidth (2025). At Q4_K_M quality or better with an 8K context it runs 55 of the open models we track; the largest that fits with room to spare is MiniMax-M3.1-Flash (MoE). The cheapest rental right now is $5.490/hr (Runpod, on demand).
- VRAM
- 288 GB
- Bandwidth
- 8000 GB/s
- Board power
- 1000 W
- Launched
- 2025
Source: AMD specification · bandwidth not yet confirmed by a second source, so we don't estimate speed from it
Rent the AMD Instinct MI350X
| Provider | Type | From, per hour | Rent |
|---|---|---|---|
| Runpod | On demand | $5.490/hr | Runpod |
Lowest price per provider in our hourly snapshot (). Rent buttons are referral links. Every card side by side: cloud GPU prices.
Before you rent the AMD Instinct MI350X: Runpod vs Vast.ai vs Lambda · is Vast.ai safe · is Runpod worth it · rent or buy · every AI GPU's buy and rent price.
AMD Instinct MI350X rental price history
The cheapest AMD Instinct MI350X offer we saw each day on the providers we track, since Sep 29. Renting it on demand costs $5.49/hr now; the lowest since Nov 11 was $5.49/hr on Nov 11 (Runpod), so today is the lowest we've seen.
- On demand now
- $5.49/hr
- Lowest on demand since Nov 11
- $5.49/hr · Nov 11
- Cheapest offer now
- $5.49/hr
- Lowest offer since Sep 29
- $5.49/hr · Sep 29
- Cheapest offer, incl. interruptible
- Cheapest on demand (from Nov 2)
Daily minimum of our hourly readings of Vast.ai, Runpod and Lium. Interruptible (spot) offers can be taken back mid-job. Before Nov 2 Vast mixed spot into on-demand and Runpod listed Community Cloud rates, so the on-demand line starts then. Every number is in our open daily price file.
Every day as a table
| Day | Cheapest offer | Cheapest on demand |
|---|---|---|
| Nov 18 (now) | $5.49/hr · Runpod | $5.49/hr · Runpod |
| Nov 17 | $5.49/hr · Runpod | $5.49/hr · Runpod |
| Nov 16 | $5.49/hr · Runpod | $5.49/hr · Runpod |
| Nov 15 | $5.49/hr · Runpod | $5.49/hr · Runpod |
| Nov 14 | $5.49/hr · Runpod | $5.49/hr · Runpod |
| Nov 13 | $5.49/hr · Runpod | $5.49/hr · Runpod |
| Nov 12 | $5.49/hr · Runpod | $5.49/hr · Runpod |
| Nov 11 | $5.49/hr · Runpod | $5.49/hr · Runpod |
| Oct 22 | $5.49/hr · Runpod | — |
| Oct 21 | $5.49/hr · Runpod | — |
| Oct 20 | $5.49/hr · Runpod | — |
| Oct 19 | $5.49/hr · Runpod | — |
| Oct 18 | $5.49/hr · Runpod | — |
| Oct 17 | $5.49/hr · Runpod | — |
| Oct 16 | $5.49/hr · Runpod | — |
| Oct 15 | $5.49/hr · Runpod | — |
| Oct 14 | $5.49/hr · Runpod | — |
| Oct 13 | $5.49/hr · Runpod | — |
| Oct 12 | $5.49/hr · Runpod | — |
| Oct 11 | $5.49/hr · Runpod | — |
| Oct 10 | $5.49/hr · Runpod | — |
| Oct 9 | $5.49/hr · Runpod | — |
| Oct 8 | $5.49/hr · Runpod | — |
| Oct 7 | $5.49/hr · Runpod | — |
| Oct 6 | $5.49/hr · Runpod | — |
| Oct 5 | $5.49/hr · Runpod | — |
| Oct 4 | $5.49/hr · Runpod | — |
| Oct 3 | $5.49/hr · Runpod | — |
| Oct 2 | $5.49/hr · Runpod | — |
| Oct 1 | $5.49/hr · Runpod | — |
| Sep 30 | $5.49/hr · Runpod | — |
| Sep 29 | $5.49/hr · Runpod | — |
Which AI models run on the AMD Instinct MI350X
Computed by our VRAM engine for an 8K context: the best quality each model fits at in 288 GB. Speeds are upper-bound estimates from memory bandwidth, not benchmarks.
| Model | Best fit | VRAM used | ~tok/s |
|---|---|---|---|
| MiniMax-M3.1-Flash (MoE) | NVFP4 | 235 GB81% | — |
| MiniMax-M3 (MoE) | Q5_K_M | 284 GB99% · tight | — |
| Qwen3 235B-A22B (MoE) | Q8_0 | 235 GB82% | — |
| MiniMax-M2.5 (MoE) | FP8 | 217 GB75% | — |
| Qwen3.8-Flash-Next (MoE) | Q8_0 | 179 GB62% | — |
| Qwen3.5 122B-A10B (MoE) | F16 | 234 GB81% | — |
| NVIDIA Nemotron 3.5 Super VL 120B-A12B (MoE) | F16 | 233 GB81% | — |
| NVIDIA Nemotron 3 Super 120B-A12B (MoE) | F16 | 231 GB80% | — |
| Mistral Small 4 119B (MoE) | FP8 | 114 GB39% | — |
| gpt-oss 120B (MoE) | MXFP4 | 62 GB22% | — |
| Llama 4 Scout 17B-16E (MoE) | F16 | 204 GB71% | — |
| GLM-4.5-Air (MoE) | F16 | 200 GB69% | — |
Too big for 288 GB (rent a bigger card instead): what it costs to run GLM-5.3 (MoE), what it costs to run DeepSeek V4.1 Flash (MoE), what it costs to run DeepSeek V4 Pro (MoE), what it costs to run Kimi K3 (MoE).
Check a model and context of your own in the VRAM calculator.
Cards to compare it with
- B300 (HGX) · 288 GB
- B200 · 192 GB
- AMD Instinct MI300X · 192 GB
- H200 SXM · 141 GB
- H200 NVL · 141 GB
- RTX PRO 6000 Blackwell Server Edition · 96 GB
Every card by VRAM tier: all GPUs for gaming and local AI.