gpt-oss-120b pricing & benchmarks
gpt-oss-120b is available from OpenAI at $0.150 per million input tokens and $0.600 per million output tokens ($0.262 blended at 3:1). It accepts up to 131,072 tokens of context. gpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and general-purpose production use cases. It activates 5.1B parameters per forward pass and is optimize…
Pricing
Per million tokens · OpenRouter
Input$0.150
Output$0.600
Blended (3:1)$0.262
Specification
openai/gpt-oss-120b
Context window131K
Max output66K
ReleasedAug 2025
LicenceOpen weights
Modalitiesin:text, out:text
Benchmarks
Independently run — not vendor self-reports. Every score links its source.
| Benchmark | Score | Config | Run | Source |
|---|---|---|---|---|
| GPQA Diamond | 75.8% ±2.7 | high effort | 2025-12-11 | Independently run · Epoch AI |
| OTIS Mock AIME | 88.9% ±4.5 | high effort | 2025-12-11 | Independently run · Epoch AI |
Scores published by Epoch AI, 'AI Benchmarking Hub', used under CC BY 4.0.
Cheaper models that score at least as well on GPQA Diamond
Strictly lower blended price AND an equal-or-higher GPQA Diamond score. One benchmark is one dimension — a model that wins here may still be weaker at your specific task.
| Model | Provider | Blended $/1M | GPQA Diamond | |
|---|---|---|---|---|
| Qwen3.7 Flash | Alibaba | $0.055 (−79%) | 82.3% | Compare → |
| Qwen3.5-Flash | Alibaba | $0.114 (−57%) | 82.3% | Compare → |
| Gemma 4 31B | $0.250 (−5%) | 75.8% | Compare → |
Where to run gpt-oss-120b
24 options across 22 providers · 7.5× between cheapest and dearest. Blended at 3:1. ⚠ marks an option serving less context than the best available, or re-pricing above a prompt-length threshold. Bracketed labels are the provider's own service tier or region.
| Host | Blended $/1M | In / Out | Context | Weights | Uptime 24h |
|---|---|---|---|---|---|
| Near AI | $0.060 | $0.030 / $0.150 | 131K | fp4 | 97.5% |
| Venice | $0.060 | $0.030 / $0.150 | 128K | int4 | 99.6% |
| CoreWeave | $0.065 | $0.030 / $0.170 | 131K | fp4 | 99.7% |
| DekaLLM | $0.068 | $0.030 / $0.180 | 131K | bf16 | 100.0% |
| AkashML | $0.069 | $0.030 / $0.187 | 131K | bf16 | 99.9% |
| DeepInfra | $0.070 | $0.037 / $0.170 | 131K | bf16 | 100.0% |
| Crusoe | $0.100 | $0.050 / $0.250 | 131K | bf16 | 99.7% |
| Mancer 2 | $0.100 | $0.050 / $0.250 | 131K | fp8 | 99.8% |
| Novita | $0.100 | $0.050 / $0.250 | 131K | fp4 | 100.0% |
| DigitalOcean | $0.150 | $0.060 / $0.420 | 128K | — | 100.0% |
| Google (global) | $0.158 | $0.090 / $0.360 | 131K | — | 87.6% |
| BaseTen (fp4/us) | $0.200 | $0.100 / $0.500 | 128K | fp4 | 99.8% |
| BaseTen | $0.200 | $0.100 / $0.500 | 128K | fp4 | 99.4% |
| Amazon Bedrock (eu-west-1) | $0.262 | $0.150 / $0.600 | 131K | — | 100.0% |
| Nebius | $0.262 | $0.150 / $0.600 | 131K | fp4 | 100.0% |
| Amazon Bedrock | $0.262 | $0.150 / $0.600 | 131K | — | 100.0% |
| SiliconFlow | $0.262 | $0.150 / $0.600 | 131K | fp8 | 95.7% |
| Phala | $0.262 | $0.150 / $0.600 | 131K | — | 100.0% |
| Together | $0.262 | $0.150 / $0.600 | 131K | — | 97.6% |
| Groq | $0.262 | $0.150 / $0.600 | 131K | — | 100.0% |
| Parasail | $0.263 | $0.100 / $0.750 | 131K | fp4 | 100.0% |
| Mara | $0.300 | $0.150 / $0.750 | 131K | — | 92.2% |
| SambaNova | $0.343 | $0.140 / $0.950 | 131K | — | 100.0% |
| Cerebras | $0.450 | $0.350 / $0.750 | 131K | fp16 | 100.0% |
Host prices and uptime from OpenRouter, synced 2026-10-22. Lower-precision weights (fp4, fp8) can cost less and score worse than the same model at bf16 — the score above is not host-specific.
Compare gpt-oss-120b
FAQ
How much does gpt-oss-120b cost?
gpt-oss-120b costs $0.150 per million input tokens and $0.600 per million output tokens on OpenRouter. At a 3:1 input:output mix that blends to $0.262 per million tokens.
What is gpt-oss-120b's context window?
gpt-oss-120b accepts up to 131,072 tokens of context, and can generate up to 65,536 output tokens.
How good is gpt-oss-120b on benchmarks?
gpt-oss-120b scores 75.8% on GPQA Diamond, independently run and published by Epoch AI. 2 benchmark results are listed on this page, each with its run date and source.
Is there a cheaper model as good as gpt-oss-120b?
Yes — 3 models in our catalogue cost less per token than gpt-oss-120b and score at least as high on the same benchmark. The cheapest is Qwen3.7 Flash at $0.055 per million tokens blended.