Checked Oct 17, 2026
OpenAI
gpt-oss-120b
Best for: Cheap or self-hosted reasoning and agents on a single GPU, when open weights matter.
- LMArena text #215 of 416
- Reasoning
- Open weights
- Reads text
- Input
- $0.15
- per 1M tokens
- Output
- $0.60
- per 1M tokens
- Context window
- 128K
- tokens in one request
- Max output
- 128K
- tokens per reply
- Released
- Aug 5, 2025
- Knowledge cutoff
- Jun 2024
- Weights
- Open
- Apache 2.0
OpenAI's open-weight gpt-oss-120b served on Together AI serverless inference (MXFP4); the price applies to Together only, other hosts differ. No cached-input price listed.
What would gpt-oss-120b cost you?
Change the numbers; the chart compares all 10 featured models at the same usage.
What it's good at
From OpenAI's own announcements and docs, linked.
- An open-weight mixture-of-experts model (117B parameters, 5.1B active per token) under the Apache 2.0 license, so you can self-host it, customise it and use it commercially. source
- OpenAI says it nearly matches o4-mini on core reasoning benchmarks, with low, medium and high reasoning levels. source
- Built for agents: function calling, web browsing, Python execution and structured outputs, and it can be fine-tuned on a single H100 node. source
Watch out for
Limits and costs worth knowing before you build on it.
- Text only (no images), and its knowledge stops at June 2024.
- Needs OpenAI's 'harmony' prompt format; most hosts handle this for you.
- Weaker factual recall: OpenAI's model card reports frequent wrong answers on SimpleQA without web browsing.
- OpenAI's own API lists no capacity for it at the standard tiers, so use a host such as Together AI or run it yourself. The price shown is Together AI's; other hosts differ.
Scores
Benchmark numbers are as published by OpenAI; we haven't rerun them. Makers use different setups, so don't compare one maker's numbers with another's. The leaderboard rank comes from public blind votes.
| Measure | Result | Source |
|---|---|---|
| LMArena text rank | #215 (1352) | Oct 15, 2026 |
| GPQA Diamond (science questions, no tools) | 80.1% | OpenAI |
| MMLU (general knowledge) | 90.0% | OpenAI |
| SWE-Bench Verified (coding) | 62.4% | OpenAI |
How to start
Get an API key
Sign up on Together AI's developer platform and add billing; you pay per token as shown above. Read the docs.
Or run it yourself
Download the weights from Hugging Face and serve them on your own or a rented GPU (see below).
Run gpt-oss-120b on your own GPU
Open weights mean you can rent a GPU by the hour and serve the model yourself: a flat hourly cost however many tokens you generate. It needs about 80 GB of GPU memory. OpenAI's 4-bit (MXFP4) release fits on a single 80 GB GPU such as an H100; the smaller gpt-oss-20b runs in 16 GB.
| Provider | From | Starting plan | Free tier | Our rating | Link |
|---|---|---|---|---|---|
| RunPodChecked | $0.34/hr | RTX 4090 24GB (Community Cloud) | No free credits; small referral bonus after a $10 top-up | Not rated | VisitOfficial site |
| TensorDockChecked (partly unverified) | $0.35/hr | RTX 4090 24GB | No ($5 minimum deposit) | Not rated | VisitOfficial site |
| Vast.aiChecked | $0.49/hr | RTX 4090 24GB | No free credits | Not rated | VisitOfficial site |
| DigitalOcean GPU DropletsChecked | $0.76/hr | RTX 4000 Ada 20GB8 vCPU · 32 GB RAM | New-account credit; GPU use not confirmed | Not rated | VisitOfficial site |
| ModalChecked | $1.10/hr | A10 | $30/month of free credits on the Starter plan | Not rated | VisitOfficial site |
| HyperstackChecked | $1.35/hr | A100 80GB (PCIe)A100 PCIe · 28 vCPU · 120 GB RAM | No | Not rated | VisitOfficial site |
| Vultr Cloud GPUChecked | $1.67/hr | L40S 48GB (1 GPU)16 vCPU · 180 GB RAM | New-customer promo credit; GPU use not confirmed | Not rated | VisitOfficial site |
| LambdaChecked | $2.29/hr | GH200 96GB (1x instance)64 vCPU · 432 GB RAM | No | Not rated | VisitOfficial site |
Compare with the other 9
| Model | Input / 1M | Output / 1M | Context | Arena |
|---|---|---|---|---|
| gpt-oss-120bOpenAI | $0.15 | $0.60 | 128K | #215 |
| Gemini 3.8 FlashGoogle | $0.75 | $3.75 | 1M | #9 |
| DeepSeek V4 ProDeepSeek | $1.32 | $3.96 | 1M | #56 |
| Grok 4.7xAI | $2 | $6 | 500K | #89 |
| Gemini 4 ArgonGoogle | $2 | $10 | 1M | #1 |
| GPT-6.1 SolOpenAI | $2 | $10 | 1.05M | #22 |
| Claude Sonnet 5.5Anthropic | $2 | $10 | 1M | #36 |
| Claude Opus 5.5Anthropic | $4 | $20 | 1M | #3 |
| Claude Fable 5.1Anthropic | $10 | $50 | 1M | #6 |
| GPT-6 AstraOpenAI | $10 | $50 | 1.05M | #30 |
Cheapest first, using 3 input tokens per output token.
Questions
How much does gpt-oss-120b cost?
gpt-oss-120b costs $0.15 per million input tokens and $0.60 per million output tokens at standard API rates through Together AI (US dollars, checked Oct 17, 2026). It is the 1st cheapest of the 10 models we feature.
How big is gpt-oss-120b's context window?
131,072 tokens for the prompt, documents and conversation in one request. A single reply can be up to 131,072 tokens. How many words that holds depends on the model's tokenizer.
Is gpt-oss-120b open source?
Its weights are published under the Apache 2.0 license, so you can download it and run it on your own GPU or a rented one, as well as use it through hosted APIs.
When was gpt-oss-120b released?
OpenAI released it in Aug 5, 2025.
What is gpt-oss-120b's knowledge cutoff?
OpenAI gives its training-data cutoff as Jun 2024. For later events it needs web search or documents you provide.
Sources
- Together AI pricing, checked Oct 17, 2026
- huggingface.co/openai/gpt-oss-120b
- openai.com/index/introducing-gpt-oss
- developers.openai.com/api/docs/models/gpt-oss-120b
- arxiv.org/abs/2508.10925
Facts checked Oct 17, 2026. Model names belong to their makers; AskGPU isn't affiliated with OpenAI. Spotted something out of date? Tell us at hello@askgpu.com.