Ultra-cheap interruptible GPUs
Batch/vision jobs that can restart; minimize $/GPU-hr on consumer or spot capacity.
At a glance — Oct 2026
Updated 2026-10-27 · forward this page: /jtbds/distributed-cheap-gpu#briefScores as of 2026-10-19 (8 days old).
- Depot Pick
- SaladCloud · 6.96/10 for this job (curated estimate, high confidence)
Pure pricePerf; reliability intentionally discounted.
- Runner-up
- Vast.ai · 7.20/10
Scores higher than the pick on: Capability fit, Price / performance, Developer & agent DX, Lock-in risk (higher = worse), Maturity / track record.
- When not to use the pick
- Need SLA, multi-node IB, or customer PII guarantees.
- Lock-in / migration
- Runner-up Vast.ai has lower stored lock-in risk (2.5 vs 3.5, 0–10, higher = more lock-in).
- Sourced price (usd_per_gpu_hour)
Pick's lowest sourced SKU: RTX 3090 24GB (Lowest priority, batch) $0.09 (usd_per_gpu_hour) — source, as of 2026-10-19
Cheapest matching SKU: SaladCloud RTX 3090 24GB (Lowest priority, batch) $0.09 — source, as of 2026-10-19. The pick is the cheapest.
Price history: first seen 2026-10-19 · last changed no change recorded · history JSON
- Free option
- No free resource lists this job. A cheap paid SKU is not a free option.
- Failure modes to watch
- Partial tool success: Require a per-item status and a failure count. Success for the task requires the failed count to be zero or explicitly handed to a retry queue.
- Runaway spend: Set a budget and a tool-call cap per task. Alert when spend jumps. Stop the run at the cap instead of summarizing a partial trace as success.
Cite: Agent Provider Depot, Depot Pick for distributed-cheap-gpu: salad, scores as of 2026-10-19. Agents: GET /api/decide?jtbd=distributed-cheap-gpu
All sourced prices, cheapest first
SaladCloud
Pure pricePerf; reliability intentionally discounted.
Pricing (directional): As of 2026-10-24 (salad.com/pricing, fetched 2026-10-24): GPU classes 'from $0.015 per GPU-hour at Lowest priority'; 'Billing is per second while an instance is in the running state.' 'Allocation, image download, and container start are fre…
Lock-in risk: higher = more lock-in (worse). Overall already inverts this dimension.
Promoted high on official /pricing floor ($0.015/GPU-hr). Depot interruptible ultra-cheap GPU pick; Vast/TensorDock remain marketplace peers. Prefer Nebius/RunPod/Modal when you need sticky sessions or predictable TTFT.
Quote for Ultra-cheap interruptible GPUs
Depot Pick stays SaladCloud. Prices below are sourced SKU rows only (as of 2026-10-27). Input and output are not blended. GET /api/quote?jtbd=distributed-cheap-gpu
Only 1 candidate has a matching USD per GPU-hour; nothing was compared, so pickIsPriceWinner is null.
- RTX 3090 24GB (High priority) usd_per_gpu_hour $0.17 source as of 2026-10-19
- RTX 3090 24GB (Lowest priority, batch) usd_per_gpu_hour $0.09 source as of 2026-10-19
- RTX 4090 24GB (High priority) usd_per_gpu_hour $0.33 source as of 2026-10-19
- RTX 4090 24GB (Lowest priority, batch) usd_per_gpu_hour $0.16 source as of 2026-10-19
- RTX 5090 32GB (High priority) usd_per_gpu_hour $0.5 source as of 2026-10-19
- RTX 5090 32GB (Lowest priority, batch) usd_per_gpu_hour $0.25 source as of 2026-10-19
3 of 4 candidates have no sourced price row. 3 are excluded from the USD per GPU-hour comparison (gpu_hour). min sourced GPU-hour SKU among candidates; not coerced to $/1M; per-second GPU was not included; not quality-adjusted
Runners-up
Vast.ai
Cheapest GPUs if your job can die and restart. Never put customer PII training here without eyes wide open.
TensorDock
Cheap GPU peer to Vast/RunPod community. Fine for interruptible training/batch; not regulated PII.
Oblivus
Price-check peer to Nebius/RunPod for on-demand H100s. Validate support and regions.
When NOT to use the top pick
Need SLA, multi-node IB, or customer PII guarantees.
Full ranking (score table)
Ranked with this JTBD's weights. Scores are curated estimates unless marked otherwise. Agents: prefer /api/decide then this table for tie-breaks.
| # | Provider | Overall | Fit | $/perf | Latency | DX | As of |
|---|---|---|---|---|---|---|---|
| 1 | Oblivus | 7.3 | 7.4 | 8.4 | 7.0 | 7.2 | 2026-09-14 |
| 2 | TensorDock | 7.2 | 7.2 | 8.8 | 6.2 | 7.2 | 2026-09-14 |
| 3 | Vast.ai | 7.2 | 7.0 | 9.8 | 4.5 | 7.2 | 2026-09-14 |
| 4 | SaladCloudDepot Pick | 7.0 | 6.5 | 9.5 | 5.0 | 7.0 | 2026-10-19 |
Score methodology for this job
Pure pricePerf; reliability intentionally discounted.
capabilityFit: 22%pricePerformance: 25%latencyReliability: 15%developerDx: 12%privacyPosture: 8%lockInRisk: 8%maturity: 10%