01
Compare approaches
Provide requirements and constraints, then ask for a recommendation with explicit trade-offs and assumptions.
gpt-oss-20b API.
Put reasoning in your app.
gpt-oss-20b is an open-weight reasoning model from OpenAI. Use Blade’s hosted endpoint to evaluate planning, analysis and coding assistance with your own product context.
€5 in free credits for every new customer.
Try gpt-oss-20bFrom €0.0183 per million input tokens. View the pricing details.
Reasoning & coding
with gpt-oss-20b.
Practical starting points for gpt-oss-20b. Choose one task, define the expected output and measure the result on your own examples.
02
Assist code review
Send a focused diff and relevant context. Ask for actionable issues, then check each finding against the actual code.
03
Build task-oriented assistants
Turn a user’s request into a proposed plan. Keep tool execution and permissions in your application rather than in the generated text.
gpt-oss-20b pricing
gpt-oss-20b
API pricing.
These are catalogue “From” rates in EUR. Dynamic pricing can change the applicable price. Confirm your model, request route and active billing unit before estimating a workload.
See gpt-oss-20b on the rate card ↗- Input / 1M tokens
- From€0.0183
- Output / 1M tokens
- From€0.0697
A concrete example
10,000 input tokens + 2,000 output tokens = €0.000322 at the displayed starting rates.
Illustration only; excludes retries, other models and paid tools. Catalogue updated 2026-10-22.
Check current rates ↗API integration
How to use
the gpt-oss-20b API.
Use the Blade model ID gpt-oss-20b and keep your API key in your backend environment. Verify account access and supported parameters before connecting production traffic.
- Create a Blade API key in the console.
- Run
pip install openaiand setBLADE_API_KEY. - Run a small request and inspect the output before building the rest of your workflow.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.models.blade.sh/v1",
api_key=os.environ["BLADE_API_KEY"],
timeout=120.0,
)
reply = client.chat.completions.create(
model="gpt-oss-20b",
messages=[{
"role": "user",
"content": "Review a retry strategy for a payment API. Explain when retries are unsafe and how idempotency helps.",
}],
max_tokens=4096,
)
print(reply.choices[0].message.content)Install openai and set BLADE_API_KEY. This sends a billed request. Check supported parameters and handle timeouts, rate limits and incomplete responses.
Use gpt-oss-20b
with Claude Code or Codex.
Connect a skill or API runner that sends a focused task to gpt-oss-20b. Your main assistant keeps the project context, reviews the proposal and runs the checks.
01 / BriefClaude Code or Codex
Define the task and select the files.
02 / Generategpt-oss-20b
Propose code through Blade’s API.
03 / CheckYour project
Review the diff. Run the tests.
A configured integration is required. Your assistant retains its main model; delegated Blade requests are billed separately. Evaluate the result on your task before expanding the workflow.
Set up coding delegation ↗gpt-oss-20b API FAQ.
What can I use gpt-oss-20b for?
gpt-oss-20b is an open-weight reasoning model from OpenAI. Use Blade’s hosted endpoint to evaluate planning, analysis and coding assistance with your own product context. Its Blade model ID is gpt-oss-20b.
How much does gpt-oss-20b cost on Blade?
Input / 1M tokens: from €0.0183. Output / 1M tokens: from €0.0697. These are dynamic catalogue starting rates as of 2026-10-22, not fixed prices for every request. Check the console for current rates.
What should I confirm before integrating?
OpenAI’s model card describes the upstream model. Blade’s alias, serving configuration and supported reasoning parameters determine what your API request can use.
Can I use gpt-oss-20b with Claude Code or Codex?
Yes, through a configured delegation skill or API runner. Your primary assistant sends a task to Blade and reviews the result. It keeps its own model, and Blade requests are billed separately. Follow the delegation guide.
Model reference: Publisher model reference ↗. Hosted configuration and pricing: Blade catalogue, 2026-10-22.
Explore related models.
Choose by task, input type and billing unit. Browse all models ↗
Need to configure your own worker? Read the gpt-oss-20b GPU deployment guide ↗