Cheapest way to run gpt-oss-120b
We earn commissions when you shop through the links below.
gpt-oss-120b needs 63.1 GB at 4-bit — cheapest rental $0.40/hr.
Quick answer
At 4-bit with an 8K-token conversation, gpt-oss-120b needs about 63.1 GB. The cheapest rental big enough is A100 PCIe on Vast.ai at $0.40 an hour; Novita AI hosts it for $0.05 per 1M input tokens.
| Option | What you need | Cost | Notes |
|---|---|---|---|
| Your own GPU | Doesn't fit any single consumer card in our list (largest 32 GB) at 4-bit with 8K context | – | – |
| Mac (unified memory) | Mac, 96 GB at MXFP4 (native), needs 65.7 GB | No Mac prices in our data | $0 per token once you own it |
| Rent a GPU | A100 PCIe (80 GB total) on Vast.ai Referral link, Marketplace, verified on-demand | $0.40 per hour | Pay only while it runs. Price checked 2026-11-09 |
| Hosted API | Novita AI: GPT OSS 120B Referral link | $0.05 in / $0.25 out per 1M tokens | Pay per token, nothing to set up. Price checked Nov 3 2026 |
Fit rule: the Q4_K_M file (or the highest precision that fits) plus an 8K-token FP16 KV cache under 90% of memory, same as the calculator. Sizes and sources: gpt-oss-120b VRAM page. Rental prices: GPU prices (checked 2026-11-09). Hourly and per-token prices measure different things, so which is cheaper depends on how much you use it.
Links marked Referral link, Affiliate partner or Amazon affiliate link may earn us credit or a commission at no cost to you. Disclosure.
Precision ladder: what each quant costs
| Precision | Needs at 8K | Smallest consumer GPU | Smallest Mac | Cheapest single-GPU rental | Cheapest rental overall (any GPU count) |
|---|---|---|---|---|---|
| MXFP4 (native) | 65.7 GB | none (max 32 GB) | Mac, 96 GB at MXFP4 (native) | A100 PCIe (80 GB) on Vast.ai, $0.40/hr | A100 PCIe (80 GB) on Vast.ai, $0.40/hr |
| 8-bit (Q8_0) | 63.7 GB | none (max 32 GB) | Mac, 96 GB at 8-bit (Q8_0) | A100 PCIe (80 GB) on Vast.ai, $0.40/hr | A100 PCIe (80 GB) on Vast.ai, $0.40/hr |
| 4-bit (Q4_K_M) | 63.1 GB | none (max 32 GB) | Mac, 96 GB at 4-bit (Q4_K_M) | A100 PCIe (80 GB) on Vast.ai, $0.40/hr | A100 PCIe (80 GB) on Vast.ai, $0.40/hr |
Every row names its precision, Mac rows included (Macs share unified memory with the system, so the same 90% rule applies). Single-GPU and multi-GPU rentals are listed separately: a multi-GPU rental needs software that splits the model across cards (llama.cpp, vLLM tensor parallel). Prices checked 2026-11-09.
How context length changes the bill (4-bit)
| Context | Needs | Cheapest rental overall |
|---|---|---|
| 8K tokens | 63.1 GB | A100 PCIe (80 GB) on Vast.ai, $0.40/hr |
| 32K tokens | 64.0 GB | A100 PCIe (80 GB) on Vast.ai, $0.40/hr |
| 128K tokens | 67.7 GB | A100 PCIe (80 GB) on Vast.ai, $0.40/hr |
What $100 of rental buys you
We list no GPU purchase prices (they change daily and we cannot check them), so this page does not compare renting with buying. At the cheapest rental that fits (A100 PCIe (80 GB) on Vast.ai, $0.40/hr), $100 of rental buys about 250 hours of running gpt-oss-120b at 4-bit with 8K context. Electricity, storage and setup time are not counted.
Fit quirks for this model
- Mixture-of-experts: about 5.1B parameters are active per token, but every expert must still be in memory, so VRAM follows the total size, not the active size.
- 18 full-attention layers × 8 KV heads × 64 head dim × 2 × 2 bytes; sliding-window layers capped at 128 tokens.
- Experts stay in MXFP4, so quant sizes are close.
- Maximum context is 131,072 tokens; longer settings are capped.
Model data checked 2026-10-23; rental prices checked 2026-11-09. For file sizes, sources and the full fit table see the gpt-oss-120b VRAM requirements page; this page is only about cost.
Get fit alerts for your GPU
Join the fit-alerts waitlist
Alerts are not running yet. Join the waitlist and pick your card once; when alerts launch we will email you once to confirm, and only after you confirm will you get the first alerts. Unsubscribe any time.
More: all cheapest-way pages · cheapest GPU rental by VRAM · Ask VRAMNeeded · Scan my GPU