OpenAI · LLM API pricing
gpt-oss-20b API pricing: cheapest provider
10 providers sell gpt-oss-20b. W&B Inference (CoreWeave) is cheapest; the most expensive charges 4.1× as much for the same model.
- Cheapest$0.055per 1M · W&B Inference (CoreWeave)
- Input · output$0.03 / $0.13per 1M tokens
- Price gap4.1×across 10 providers
- Accuracy60.8%GPQA · medium reasoning
- Context131Kmax tokens
Standard list prices in USD from each provider's pricing page · last price change Oct 17, 2026 · How we collect and rank prices
Ranked for your volume
gpt-oss-20b monthly cost by provider
Change the monthly tokens and the ranking updates instantly, using each provider's standard list price.
More requirements: caching, context, features, provider
| Provider | Estimated / month | Input / 1M | Output / 1M | Cached input / 1M | Context |
|---|---|---|---|---|---|
1 W&B Inference (CoreWeave)Inference provider - lowest for your volumePricing page | $0.56 | $0.03 | $0.13 | Not listed | 131,000 |
2 DeepInfraInference providerPricing page | $0.58 | $0.03 | $0.14 | Not listed | 131,072 |
3 Novita AIInference providerPricing page | $0.70 | $0.04 | $0.15 | Not listed | 131,072 |
4 SiliconFlowInference providerPricing page | $0.76 | $0.04 | $0.18 | Not listed | 131,000 |
5 OVHcloud AI EndpointsCloud platformBilled in EUR: €0.04 input, €0.15 output per 1M tokens. USD at the ECB rate of Oct 30, 2026 (1.1172).Pricing page | $0.78 | $0.0447 | $0.1676 | Not listed | 131,000 |
6 Amazon BedrockCloud platformPricing page | $1.30 | $0.07 | $0.30 | Not listed | 131,072 |
7 Fireworks AIInference providerPricing page | $1.30 | $0.07 | $0.30 | $0.035 | 131,072 |
8 Oracle Cloud Generative AICloud platformPricing page | $1.30 | $0.07 | $0.30 | Not listed | 128,000 |
9 ReplicateInference providerPricing page | $1.62 | $0.09 | $0.36 | Not listed | Not listed |
10 Cloudflare Workers AICloud platformPricing page | $2.60 | $0.20 | $0.30 | Not listed | 128,000 |
Order depends on listed price for your volume only. No provider pays for its position.
Input volume includes the cached share, not an additional token volume. This is a list-price estimate, not a quote: cache-write charges, cache eligibility/TTL, batch discounts, long-context surcharges, regional pricing and taxes are excluded. Check the provider's terms for your workload before switching.
Accuracy: GPQA Diamond (graduate-level science questions) measured by Epoch AI (CC BY 4.0) on gpt-oss-20b_medium, run Aug 27, 2026. The model is the same at every provider, though quantized deployments can score lower.
Price history
The cheapest price has not changed since tracking started on Sep 28, 2026: $0.055 per 1M tokens at W&B Inference (CoreWeave). All price changes
Switch to the cheapest provider
Already using the OpenAI SDK? Change the base URL, key and model name:
from openai import OpenAI
import os
client = OpenAI(
base_url="https://api.inference.wandb.ai/v1",
api_key=os.environ["WANDB_API_KEY"],
)
response = client.chat.completions.create(
model="openai/gpt-oss-20b",
messages=[{"role": "user", "content": "Hello"}],
)Compare gpt-oss-20b with another model
Or pick any model
Before you pick a provider
Context windows differ by provider, from 128K to 131K tokens. A cheaper host may not accept your longest prompts.
1 of 10 providers list a cached-input price. Repeated system prompts or documents can cost much less there than the standard input rate shown above.
Keep comparing
More OpenAI models
gpt-oss-120b, GPT-5, GPT-5 Mini, GPT-5 Nano, GPT-4.1, GPT-4.1 mini, GPT-4o, GPT-4o mini
All OpenAI API prices - All LLM API prices - Watch gpt-oss-20b prices (RSS) - Methodology