OpenAI: gpt-oss-20b

OpenAI: gpt-oss-20b is a large language model from OpenAI. It costs $0.030 per million input tokens and $0.130 per million output tokens. Its context window is 131K tokens.

ReasoningTool callingStructured outputPrompt cachingBatch tierOpen weights

Input / 1M tokens: $0.030
Output / 1M tokens: $0.130
Cached input / 1M: $0.030
Context window: 131K

Who serves it cheapest

13 hosts serve gpt-oss-20b. Same weights, same API - the price difference is pure margin and routing.

Providers serving OpenAI: gpt-oss-20b, cheapest first
Provider	Input / 1M	Output / 1M	Context	Throughput	Uptime 24h
DarkbloomCheapestfp8	$0.020	$0.100	131K	-	99.2%
AkashMLfp4	$0.020	$0.100	131K	-	99.7%
CoreWeavefp4	$0.030	$0.130	131K	-	100.0%
DekaLLMbf16	$0.029	$0.140	131K	-	99.7%
DeepInfrabf16	$0.030	$0.140	131K	-	100.0%
Parasailfp4	$0.030	$0.150	131K	-	99.9%
Phala	$0.040	$0.150	131K	-	99.6%
Novitafp4	$0.040	$0.150	131K	-	99.9%
SiliconFlowfp8	$0.040	$0.180	131K	-	98.1%
Together	$0.050	$0.200	131K	-	94.3%
Amazon Bedrock	$0.070	$0.150	131K	-	99.9%
Google	$0.070	$0.250	131K	-	99.9%
Groq	$0.075	$0.300	131K	-	99.8%

The spread between Darkbloom and Groq is 3.3× for identical weights. Quantization and context limits differ, so check both columns before switching.

Benchmarks

Independent scores published alongside the catalogue.

Intelligence index: 9.0
Coding index: 20.7
Agentic index: 1.4

About gpt-oss-20b

gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for...

Specifications

OpenAI: gpt-oss-20b specifications
Model ID	`openai/gpt-oss-20b`
Provider	OpenAI
Context window	131K tokens
Max output	118K tokens
Input modalities	text
Output modalities	text
Knowledge cutoff	2024-06-30
Open weights	Yes - openai/gpt-oss-20b
Released	August 5, 2025

Cheaper alternatives

Models that cost less than gpt-oss-20b while keeping at least half its context window and every input modality it supports.

Granite 4.0 Micro

$0.017 in · $0.112 out

1.3× cheaper

Ling 3.0 Flash

$0.021 in · $0.063 out

1.7× cheaper

Mistral Nemo

$0.019 in · $0.030 out

2.5× cheaper

Nex-N2.5-Mini (free)

Free in · Free out

Nex-N2.5-Pro (free)

Free in · Free out

Frequently asked

How much does OpenAI: gpt-oss-20b cost?

$0.030 per million input tokens and $0.130 per million output tokens. Cached input reads cost $0.030 per million tokens.

What is the context window of OpenAI: gpt-oss-20b?

131K tokens, with up to 118K tokens of output per request.

Which provider serves OpenAI: gpt-oss-20b cheapest?

Darkbloom at $0.020 per million input tokens - 3.3× cheaper than Groq, the most expensive of the 13 hosts serving it.

Confirm against the source: OpenAI official pricing.