OpenAIopen weightsreasoninggpt-oss-120b pricing
$0.15 per 1M input tokens and $0.60 per 1M output tokens. In real work that is 0.04¢ per chat answer and 0.23¢ to summarize a 20-page PDF.
What gpt-oss-120b costs per task
Token counts estimated from a measured model of the same lab (GPT-6.1 Sol), rescaled by tokenizer. Reasoning tokens are billed as output.
| Task | Cost | Tokens used | Price rank | Basis |
|---|---|---|---|---|
| Answer a chat question | 0.04¢$0.40 / 1,000 | 45 in · 560 out · 103 reasoning | 47 of 336 | estimated |
| Reply to a customer email | 0.01¢$0.11 / 1,000 | 258 in · 126 out | 29 of 336 | estimated |
| Write a 1,500-word blog post | 0.14¢$1.38 / 1,000 | 59 in · 2,116 out · 172 reasoning | 34 of 336 | estimated |
| Write 10 product descriptions | 0.07¢$0.75 / 1,000 | 283 in · 1,103 out · 72 reasoning | 26 of 336 | estimated |
| Translate 1,000 words | 0.10¢$1.00 / 1,000 | 1,330 in · 1,289 out · 44 reasoning | 63 of 336 | estimated |
| Proofread one page (500 words) | 0.05¢$0.50 / 1,000 | 718 in · 662 out | 53 of 336 | estimated |
| Summarize a 20-page PDF | 0.23¢$2.25 / 1,000 | 12,987 in · 482 out · 26 reasoning | 74 of 336 | estimated |
| Extract data from an invoice | 0.03¢$0.33 / 1,000 | 373 in · 450 out | 28 of 336 | estimated |
| Analyze a 200-row spreadsheet | 0.52¢$5.20 / 1,000 | 4,779 in · 919 out · 6,553 reasoning | 70 of 336 | estimated |
| Answer from 5 retrieved documents (RAG) | 0.08¢$0.82 / 1,000 | 3,690 in · 442 out | 40 of 336 | estimated |
| Classify 100 support tickets | 0.05¢$0.50 / 1,000 | 1,261 in · 442 out · 77 reasoning | 8 of 336 | estimated |
| Review 500 lines of code | 0.31¢$3.07 / 1,000 | 4,557 in · 1,135 out · 2,843 reasoning | 36 of 336 | estimated |
| Write a SQL query | 0.03¢$0.29 / 1,000 | 126 in · 312 out · 147 reasoning | 15 of 336 | estimated |
| Run a 50-step coding agent | 18¢$183 / 1,000 | 2,237,500 in · 15,000 out | 90 of 336 | modeled |
Price at every provider
List prices per 1M tokens (input / output) at each host serving gpt-oss-120b through OpenRouter today. Hosts can differ in quantization, context and speed.
| Provider | Input / output | Cached input | Context | Uptime |
|---|---|---|---|---|
| N Near AIfp4 | $0.03 / $0.15 | $0.015 | 131K | 100% |
| V Veniceint4 | $0.03 / $0.15 | $0.03 | 128K | 100% |
| C CoreWeavefp4 | $0.03 / $0.17 | $0.03 | 131K | 100% |
 DeepInfrabf16 | $0.037 / $0.17 | - | 131K | 100% |
| D DekaLLMbf16 | $0.03 / $0.18 | $0.03 | 131K | 100% |
| A AkashMLbf16 | $0.03 / $0.19 | $0.03 | 131K | 100% |
| C Crusoebf16 | $0.05 / $0.25 | $0.05 | 131K | 100% |
| M Mancer 2fp8 | $0.05 / $0.25 | - | 131K | 100% |
 Novitafp4 | $0.05 / $0.25 | - | 131K | 100% |
 Googleunknown | $0.09 / $0.36 | - | 131K | 0% |
| D DigitalOceanunknown | $0.06 / $0.42 | $0.012 | 128K | 100% |
| B BaseTenfp4 | $0.10 / $0.50 | $0.10 | 128K | 100% |
| B BaseTenfp4 | $0.10 / $0.50 | $0.10 | 128K | 100% |
 Amazon Bedrockunknown | $0.15 / $0.60 | - | 131K | |
 Nebiusfp4 | $0.15 / $0.60 | - | 131K | 100% |
Amazon Bedrockunknown | $0.15 / $0.60 | - | 131K | 100% |
 SiliconFlowfp8 | $0.15 / $0.60 | $0.075 | 131K | 100% |
| P Phalaunknown | $0.15 / $0.60 | - | 131K | 100% |
 Togetherunknown | $0.15 / $0.60 | - | 131K | 55% |
 Groqunknown | $0.15 / $0.60 | $0.075 | 131K | 100% |
 Parasailfp4 | $0.10 / $0.75 | $0.055 | 131K | 100% |
| M Maraunknown | $0.15 / $0.75 | - | 131K | 100% |
 Cerebrasfp16 | $0.35 / $0.75 | $0.35 | 131K | 100% |
 SambaNovaunknown | $0.14 / $0.95 | - | 131K | 100% |
Price history
First seen in our price history on Aug 12, 2025 at $0.073 input / $0.29 output. Since then the output price moved from $0.29 to $0.60 per 1M tokens (17 changes).
FAQ
How much does gpt-oss-120b cost?
gpt-oss-120b costs $0.15 per 1 million input tokens and $0.60 per 1 million output tokens, with cached input at $0.075 (list price on Oct 18, 2026). In practice one chat answer costs about 0.04¢ and summarizing a 20-page PDF about 0.23¢.
Where is gpt-oss-120b cheapest?
Among 24 providers we track, the lowest list price today is at Near AI: $0.03 input and $0.15 output per 1M tokens.
Does gpt-oss-120b charge for reasoning tokens?
Yes. Reasoning (thinking) tokens are billed as output tokens. That is why the cost per task can be much higher than the visible answer suggests - our task costs include them.
Is the gpt-oss-120b cost per task measured or estimated?
Estimated: gpt-oss-120b was not run directly. Tokens come from GPT-6.1 Sol and are rescaled by the tokenizer of OpenAI. Real usage can differ.