impulse

gpt-oss-20b

Open weights

gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license.

API id gpt-oss-20bReleased 5 Aug 2025 (Aug 2025)
Input / 1M
$0.03
at W&B Inference
Output / 1M
$0.13
at W&B Inference
Context window
131K
tokens
Max output
131K
tokens per reply
Hosts
18
27 listings
Released
Aug 2025
5 Aug 2025

Pricing

List prices at W&B Inference, in US dollars per million tokens.

TierInput / 1MOutput / 1M
Standard$0.03$0.13

What it costs a month

At W&B Inference list prices, 30 days.

  • Chat assistant5,000 requests a day, 2K in / 0.5K out$18.75
  • RAG over documents2,000 requests a day, 8K in / 0.6K out$19.08
  • Coding agent1,000 requests a day, 40K in / 2K out$43.80
  • Overnight batch20,000 requests a day, 4K in / 0.4K out$103
Use your own numbers

Where to run it

18 hosts serve gpt-oss-20b. The comparison is on blended price, 3 parts input to 1 part output.

HostIn / out per 1M
gpt-oss-20b
$0.0145 / $0.07
−48% vs headline
W&B InferenceInference
gpt-oss-20b
$0.03 / $0.13
DeepInfraInference
gpt-oss-20b
$0.03 / $0.14
+5% vs headline
gpt-oss-20b
$0.04 / $0.15
+23% vs headline
GMI CloudInference
gpt-oss-20b
$0.04 / $0.15
+23% vs headline
Novita AIInference
gpt-oss-20b
$0.04 / $0.15
+23% vs headline
gpt-oss-20b
$0.04 / $0.19
+41% vs headline
Together AIInference
gpt-oss-20b
$0.05 / $0.20
+59% vs headline
gpt-oss-20b-maas
$0.07 / $0.25
+109% vs headline
gpt-oss-20b
$0.07 / $0.28
+123% vs headline
openai.gpt-oss-20b-1:0
$0.07 / $0.30
+132% vs headline
openai.gpt-oss-20b
$0.07 / $0.30
+132% vs headline
Fireworks AIInference
gpt-oss-20b
$0.07 / $0.30
+132% vs headline
databricks-gpt-oss-20b
$0.07 / $0.30
+132% vs headline
GroqInference
gpt-oss-20b
$0.075 / $0.30
+139% vs headline
ReplicateInference
gpt-oss-20b
$0.09 / $0.36
+186% vs headline
gpt-oss-20b
$0.20 / $0.30
+309% vs headline
NVIDIA NIMInference
gpt-oss-20b
-
gpt-oss-20b-mxfp4-GGUF
-
Show 8 more listings: regions, snapshots and variants
Inference APIs
Alias
gpt-oss-20b
$0.04 / $0.19
+41% vs headline
Amazon Bedrock
Mantle
openai.gpt-oss-20b
$0.07 / $0.30
+132% vs headline
us-gov-west-1 · Mantle
openai.gpt-oss-20b
$0.084 / $0.36
+178% vs headline
us-gov-east-1 · Mantle
openai.gpt-oss-20b
$0.084 / $0.36
+178% vs headline
us-gov-west-1
openai.gpt-oss-20b-1:0
$0.084 / $0.36
+178% vs headline
us-gov-east-1
openai.gpt-oss-20b-1:0
$0.084 / $0.36
+178% vs headline
us-gov
us-gov.openai.gpt-oss-20b-1:0
$0.084 / $0.36
+178% vs headline
Fireworks AI
Alias
gpt-oss-20b
$0.07 / $0.30
+132% vs headline

Capabilities

Core

  • Streaming
  • Function calling
  • JSON mode
  • Structured outputs

Multimodal

  • Vision
  • Image input
  • Audio
  • Image generation
  • Video

Advanced

  • Reasoning
  • Tool calling
  • MCP
  • Prompt caching
  • Batch API
  • Embeddings
Takes in
Text
Produces
Text
Open weights
License: Apache-2.0
Runs in about 50.2 GB of GPU memory

Cheaper alternatives

Current models that cost less and keep most of what this one can do: tools, vision, reasoning, structured output, caching and enough context.

Same model, cheaper host

  • Darkbloom
    gpt-oss-20b
    $0.0145 / $0.07
    48% less

Smaller OpenAI models

No cheaper OpenAI model keeps its key capabilities.

Other frontier labs

No cheaper match from another frontier lab.

Open-weight picks

  • Gemma 4 E2B
    Google · on Amazon Bedrock
    $0.04 / $0.08
    Has every key capability
    9% less
  • Qwen3.7 Flash Variant1
    Alibaba Qwen · on EmpirioLabs AI
    $0.028 / $0.11
    Has every key capability
    12% less
  • Qwen3.5 4B
    Alibaba Qwen · on EmpirioLabs AI
    $0.04 / $0.07
    Has every key capability
    14% less
  • Gemma 4 E4B IT
    Google · on DeepInfra
    $0.02 / $0.10
    Has every key capability
    27% less

Call it

Using W&B Inference's API.

Python
from litellm import completion

response = completion(
    model="wandb/gpt-oss-20b",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Quick answers

How much does gpt-oss-20b cost?

On W&B Inference, gpt-oss-20b costs $0.03 per million input tokens and $0.13 per million output tokens. The cheapest host is Darkbloom, at $0.0145 per million input tokens and $0.07 per million output tokens.

Which APIs offer gpt-oss-20b?

gpt-oss-20b runs on 18 APIs: Darkbloom, W&B Inference, DeepInfra, OVHcloud AI Endpoints, GMI Cloud, Novita AI, Inference APIs, Together AI, Vertex AI and 9 more.

What is gpt-oss-20b's context window?

131K tokens, with up to 131K tokens in one reply.

Is gpt-oss-20b being retired?

No retirement date is scheduled for it on any API we track.

What are cheaper alternatives to gpt-oss-20b?

Models that cost less and keep most of what it can do: Gemma 4 E2B on Amazon Bedrock (9% less). gpt-oss-20b itself costs 48% less on Darkbloom. Savings are on a mix of 3 input tokens to 1 output token.

When was gpt-oss-20b released?

OpenAI released it on 5 Aug 2025.

What is gpt-oss-20b's API model id?

Use gpt-oss-20b with W&B Inference. Other hosts use their own ids: gpt-oss-20b-maas on Vertex AI, openai.gpt-oss-20b-1:0 on Amazon Bedrock, openai.gpt-oss-20b on Oracle OCI, databricks-gpt-oss-20b on Databricks.

Sources

Raw catalog entry for gpt-oss-20b on W&B Inference
{
  "supports_reasoning": true,
  "max_input_tokens": 131000,
  "max_output_tokens": 131072,
  "input_cost_per_token": 3e-8,
  "output_cost_per_token": 1.3e-7,
  "litellm_provider": "wandb",
  "mode": "chat",
  "source": "https://wandb.ai/site/pricing/tokens/"
}