BroadRouter

gpt-oss-20b vs Gemma 3 4B API pricing

Updated

gpt-oss-20b: USD 0.018 input, USD 0.09 output per 1M tokens (AkashML, promotional). Gemma 3 4B: USD 0.04 input, USD 0.08 output per 1M tokens (Amazon Bedrock). At 3 input tokens per output token, Gemma 3 4B costs 39% more than gpt-oss-20b (USD 0.05 vs USD 0.036 per 1M).

Picked by the same-tier rule: both are in the first quarter of the indexed models by price (below USD 0.1556 per 1M tokens at 3:1) and in the same size class, small (a tier word such as Flash, Mini or Haiku, or 40B parameters or fewer). Gemma 3 4B is the 1st-closest-priced model from another maker to gpt-oss-20b in that bucket. Price only: every number is computed from prices read from each provider's own pricing page (linked per cell, newest check 2026-11-15), with no claim about quality, speed or accuracy.

How do gpt-oss-20b and Gemma 3 4B differ on price?

gpt-oss-20b and Gemma 3 4B prices side by side

USD per 1M tokens, standard tier. "Cheapest" is the lowest price at 3 input tokens to 1 output token; each line can come from a different provider.

gpt-oss-20b and Gemma 3 4B API prices compared
Pricegpt-oss-20bGemma 3 4B
Model pageGPT OSS 20B
OpenAI · 10 providers
Gemma 3 4B
Google · 3 providers
Cheapest standard price, blended 3:1USD 0.036 lower promo
USD 0.018 in / USD 0.09 out at AkashML
AkashML page checked 2026-11-15
USD 0.05
USD 0.04 in / USD 0.08 out at Amazon Bedrock
Amazon Bedrock page checked 2026-11-14
Cheapest inputUSD 0.018 lower promo
AkashML source
USD 0.04
Amazon Bedrock source
Cheapest cached inputUSD 0.007
Google Vertex AI source
No standard price lists a cache price; cached tokens bill at the input price.
Cheapest outputUSD 0.09 promo
AkashML source
USD 0.08 lower
Amazon Bedrock source
Cheapest non-promotional priceUSD 0.03 / USD 0.13
CoreWeave Forge, USD 0.055 blended source
The cheapest price is not promotional.
Maker's own APIOpenAI publishes no paid standard price on its own API that we found.Google publishes no paid standard price on its own API that we found.
Median standard priceUSD 0.04 / USD 0.19
input / output, 10 prices
USD 0.05 / USD 0.10
input / output, 3 prices
Context window131,072 tokens, largest stated131,072 tokens, largest stated; from 32,768 at some providers
Providers with a standard price103

Report a wrong price or number (email to info@broadnet.ai, with this page's link filled in).

What do gpt-oss-20b and Gemma 3 4B cost for a real workload?

30 days at each model's cheapest standard price for the workload (presets of the cost calculator; cached input at the cached price where listed).

Monthly cost of three workloads on gpt-oss-20b and Gemma 3 4B
Workloadgpt-oss-20bGemma 3 4BDifference
Support chatbot
2,000 requests a day, 1,500 input tokens (60% cached), 300 output
USD 2.75
AkashML
USD 5.04
Amazon Bedrock
USD 2.29
gpt-oss-20b lower
RAG over documents
1,000 requests a day, 8,000 input tokens (30% cached), 500 output
USD 5.02
AkashML
USD 10.80
Amazon Bedrock
USD 5.78
gpt-oss-20b lower
Coding agent (one developer)
1,500 requests a day, 40,000 input tokens (85% cached), 2,000 output incl. reasoning
USD 26.73
AkashML
USD 79.20
Amazon Bedrock
USD 52.47
gpt-oss-20b lower

gpt-oss-20b is cheaper on all three workloads. The widest gap is on Coding agent (one developer): the dearer model costs 2.96× as much as the cheaper one.

Does the cheaper model change with your input:output ratio?

The cheaper model changes with the mix: below 0.45 input tokens per output token, Gemma 3 4B is cheaper; above 0.45, gpt-oss-20b is cheaper. At the site's 3:1 blend, gpt-oss-20b is cheaper. More input per output token, as with retrieval or agents that resend long contexts, favours the lower input price; long answers favour the lower output price.

Price per 1M tokens of gpt-oss-20b and Gemma 3 4B at different input:output mixes
Input : output tokensgpt-oss-20bGemma 3 4B
1 : 3 (long answers)USD 0.072
AkashML
USD 0.07 lower
Amazon Bedrock
1 : 1USD 0.054 lower
AkashML
USD 0.06
Amazon Bedrock
3 : 1 (the site’s blend)USD 0.036 lower
AkashML
USD 0.05
Amazon Bedrock
10 : 1 (retrieval, long prompts)USD 0.02455 lower
AkashML
USD 0.04364
Amazon Bedrock
30 : 1 (agents, big contexts)USD 0.02032 lower
AkashML
USD 0.04129
Amazon Bedrock

USD per 1M tokens at the cheapest standard provider for each mix, before cache discounts. Thinking tokens bill as output.

What do gpt-oss-20b and Gemma 3 4B cost per character in other languages?

Cost to send 1M characters at each model's cheapest non-promotional standard input price, with the gpt-oss (o200k_harmony) tokenizer for gpt-oss-20b and the Gemma 3 27B tokenizer for Gemma 3 4B:

Input cost per 1M characters by language, gpt-oss-20b and Gemma 3 4B
Contentgpt-oss-20bGemma 3 4B
EnglishUSD 0.0064 lower
CoreWeave Forge
USD 0.0087
Amazon Bedrock
Arabic MSAUSD 0.0096 lower
CoreWeave Forge
USD 0.014
Amazon Bedrock
Egyptian ArabicUSD 0.0100 lower
CoreWeave Forge
USD 0.014
Amazon Bedrock
HindiUSD 0.0097 lower
CoreWeave Forge
USD 0.011
Amazon Bedrock
ChineseUSD 0.023 lower
CoreWeave Forge
USD 0.027
Amazon Bedrock
SpanishUSD 0.0068 lower
CoreWeave Forge
USD 0.0090
Amazon Bedrock
CodeUSD 0.0076 lower
CoreWeave Forge
USD 0.012
Amazon Bedrock

From the real cost by language dataset (measured tokens per character on FLORES-200 and other texts). Your own text will differ.

Check it on your own traffic

More comparisons: gpt-oss-120b vs gpt-oss-20b · Gemma 4 26B A4B vs gpt-oss-20b · gpt-oss-20b vs Qwen3.6 35B A3B · Gemma 3 12B vs Gemma 3 4B · every gpt-oss-20b comparison · every Gemma 3 4B comparison · all 662 price comparisons

Questions

Is gpt-oss-20b cheaper than Gemma 3 4B?

Yes. On a blend of 3 input tokens to 1 output token, gpt-oss-20b costs USD 0.036 per 1M tokens and Gemma 3 4B USD 0.05, at each model's cheapest standard price. For input alone, gpt-oss-20b is cheaper (USD 0.018 against USD 0.04 per 1M); for output alone, Gemma 3 4B is (USD 0.08 against USD 0.09). This compares price only, not quality.

What do gpt-oss-20b and Gemma 3 4B cost per 1M tokens?

gpt-oss-20b: from USD 0.018 input and USD 0.09 output at AkashML (promotional); median of 10 standard prices USD 0.04 / USD 0.19. Gemma 3 4B: from USD 0.04 input and USD 0.08 output at Amazon Bedrock; median of 3 standard prices USD 0.05 / USD 0.10. Each price is read from the provider's own pricing page; the source is linked in the table.

Which providers sell both gpt-oss-20b and Gemma 3 4B?

3 providers have a standard price for both: DeepInfra (USD 0.0575 and USD 0.0625 at 3:1), Featherless (USD 0.0675 and USD 0.0625 at 3:1), Amazon Bedrock (USD 0.1275 and USD 0.05 at 3:1). The first is the lowest combined price. Buying each model from its own cheapest provider costs USD 0.036 and USD 0.05.

Which is cheaper for a coding agent or a chatbot, gpt-oss-20b or Gemma 3 4B?

For a coding agent (1,500 requests a day, 40,000 input tokens (85% cached), 2,000 output incl. reasoning): USD 26.73 a month on gpt-oss-20b (AkashML) against USD 79.20 on Gemma 3 4B (Amazon Bedrock), a difference of USD 52.47. For a support chatbot (2,000 requests a day, 1,500 input tokens (60% cached), 300 output): USD 2.75 a month on gpt-oss-20b (AkashML) against USD 5.04 on Gemma 3 4B (Amazon Bedrock), a difference of USD 2.29. Each uses the cheapest standard price for that workload, cached input included. Try your own numbers.

At what input:output ratio does the cheaper of gpt-oss-20b and Gemma 3 4B change?

The cheaper model changes with the mix: below 0.45 input tokens per output token, Gemma 3 4B is cheaper; above 0.45, gpt-oss-20b is cheaper. At the site's 3:1 blend, gpt-oss-20b is cheaper. More input per output token, as with retrieval or agents that resend long contexts, favours the lower input price; long answers favour the lower output price.

Do gpt-oss-20b and Gemma 3 4B have cached input pricing?

gpt-oss-20b from USD 0.007 per 1M cached input tokens (Google Vertex AI; 4 of 10 standard prices list a discount). Gemma 3 4B: no standard price lists a cached-input discount. Cached tokens are billed at the input price where no cache price is published.

Which has the larger context window, gpt-oss-20b or Gemma 3 4B?

Both go up to 131,072 tokens on the providers that state it.

Read as MarkdownAsk ChatGPTAsk ClaudeAsk Perplexity