What is GPT-OSS-20B good for?

Local, edge and low-cost text work under Apache 2.0, small enough for a single consumer GPU.

When is GPT-OSS-20B not a good fit?

Image input, since it is text only. Its knowledge is older than the rest of OpenAI's lineup, and GPT-OSS-120B carries far more parameters under the same licence.

What is the cheapest way to run GPT-OSS-20B?

Hosted, unless you push serious volume. Geodd charges $0.03 in / $0.14 out per 1M tokens. The cheapest rental that fits is 1x RTX 4060 Ti at $79 a month, which costs the same as about 2B tokens a month on that API.

Can I self-host GPT-OSS-20B?

Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 14 GB of GPU memory, which starts at roughly $79 a month on the cheapest rental that fits.

Open weights · Apache 2.0

GPT-OSS-20B

The smaller sibling of GPT-OSS-120B, released August 2025: a 21B-parameter MoE with 3.6B active, open-weight under Apache 2.0 and designed for edge deployment on 16 GB devices.

Jump to: Specs Pricing Self-hosting Similar models FAQ More from OpenAI

Visit website Compare vs other LLMs →

Cheapest via Geodd

Input: $0.03 /1M tokens
Output: $0.14 /1M tokens

7 providers

Key Specifications

Context window: 131K tokens
Max output: 131K tokens
Knowledge cutoff: Jun 2024
Released: Aug 2025
Parameters: 21B, 3.6B active
Inputs: Text

Capabilities Show details: Reasoning Thinks before it answers, always on or as a switchable mode. Function calling Connect to external tools, APIs, and systems. Structured output Return responses in structured formats like JSON.

Hosted API pricing

Input tokens M Output tokens M

Provider	Input /1M tokens	Output /1M tokens	Cached input /1M	Cost at 10M in + 2M out
Geodd Cheapest	$0.03	$0.14		$0.58	View
Novita	$0.04	$0.15		$0.70	View
Together AI	$0.05	$0.20		$0.90	View
DigitalOcean	$0.05	$0.45		$1.40	View
Replicate	$0.09	$0.36		$1.62	View
EcoHash	$0.20	$0.28	$0.10	$2.56	View
Cloudflare	$0.20	$0.30		$2.60	View

Heads up: Base-tier, on-demand rates per 1M tokens; cached, batch and long-context tiers excluded. A provider may serve a shorter context or a quantized build than the creator's release. Verify before provisioning. More on how we price.

Estimated cost to self-host

GPT-OSS-20B needs about 14 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 1× RTX 4060 Ti at about $79 a month.

That costs the same as roughly 2B tokens a month on Geodd's API. Self-hosting is more expensive below that volume.

Fits on 1 Single node ≤ 8

Precision	Memory	Cheapest fit (32K context)(Full 131K context)	Cost /mo	Break-even vs API
4-bitINT4 / FP4	14 GB17 GB	1× RTX 4060 Ti	$79$86	2B tokens /mo2B tokens /mo
8-bitFP8 / INT8	23 GB25 GB	2× RTX 4060 Ti	$158$158	3B tokens /mo3B tokens /mo
16-bitFP16 / BF16	45 GB47 GB	4× RTX 4060 Ti	$317$317	7B tokens /mo7B tokens /mo

Estimates based on median on-demand rates for Nvidia GPUs. Memory is weights plus KV cache for one request, using FP8 KV cache where supported and FP16/BF16 otherwise. Break-even assumes a 5:1 input-to-output ratio. No guarantee of runtime support, usable performance, or that a matching quantized build exists. How we estimate costs.

Similarly priced models

The models nearest GPT-OSS-20B by blended rate, each at its own cheapest provider.

Model	Blended /1M	Input /1M	Output /1M	Context	Cutoff	vs GPT-OSS-20B
Llama 3.1 8B Meta	$0.025	$0.02	$0.05	131K	Dec 2023	−48%
DeepSeek OCR-2 DeepSeek	$0.03	$0.03	$0.03	8K		−38%
Qwen3.7-Flash Alibaba Cloud	$0.0467	$0.03	$0.13	1M		−3%
GPT-OSS-20B This model OpenAI	$0.0483	$0.03	$0.14	131K	Jun 2024
Command R7B Cohere	$0.0583	$0.04	$0.15	128K		+21%
Gemma 3 12B Instruct Google Cloud	$0.0583	$0.05	$0.10	128K	Aug 2024	+21%
Llama 3.2 1B Meta	$0.0583	$0.03	$0.20	128K	Dec 2023	+21%
GPT-OSS-120B OpenAI	$0.0633	$0.04	$0.18	131K	Jun 2024	+31%
Nemotron 3 Nano 30B A3B Nvidia	$0.075	$0.05	$0.20	262K	Jun 2025	+55%
Llama 3 8B Meta	$0.0833	$0.05	$0.25	8K	Mar 2023	+72%
Qwen3-Coder-30B-A3B Alibaba Cloud	$0.0917	$0.06	$0.25	262K		+90%

Prices are USD per 1M tokens at each model's cheapest listed provider. Blended is the cost of 10M input plus 2M output tokens, spread over the 12M.

Vast.ai

Per-second billing. Pay only for what you use.

50+ GPU models from RTX 3060 to B200, from $0.04/hr

No minimums, no long-term contracts.

SOC2 certified · PyTorch, CUDA & TensorFlow templates ready

Learn more

Frequently Asked Questions

What is GPT-OSS-20B good for?: Local, edge and low-cost text work under Apache 2.0, small enough for a single consumer GPU.
When is GPT-OSS-20B not a good fit?: Image input, since it is text only. Its knowledge is older than the rest of OpenAI's lineup, and GPT-OSS-120B carries far more parameters under the same licence.
What is the cheapest way to run GPT-OSS-20B?: Hosted, unless you push serious volume. Geodd charges $0.03 in / $0.14 out per 1M tokens. The cheapest rental that fits is 1x RTX 4060 Ti at $79 a month, which costs the same as about 2B tokens a month on that API.
Can I self-host GPT-OSS-20B?: Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 14 GB of GPU memory, which starts at roughly $79 a month on the cheapest rental that fits.

More from OpenAI

Model	Context	Cutoff	Input /1M	Output /1M
GPT-OSS-120B Open weights	131K	Jun 2024	$0.04	$0.18
GPT-5 Nano	400K	May 2024	$0.05	$0.40
GPT-4.1 Nano	1M	Jun 2024	$0.10	$0.40
GPT-4o Mini	128K	Oct 2023	$0.15	$0.60
GPT-5.6 Luna	1M	Feb 2026	$0.20	$1.20
GPT-5.4 Nano	400K	Aug 2025	$0.20	$1.25
GPT-5 Mini	400K	May 2024	$0.25	$2.00
GPT-4.1 Mini	1M	Jun 2024	$0.40	$1.60