GPT-OSS 20B Environmental Impact
Compact open-weight model — runs on 16GB edge devices
The figures below are for one quick question. A longer chat with follow-ups uses about 0.06 Wh, and an hour or two of agent work about 1.8 Wh.
One query, in context
Energy, carbon and water for a single use. Quick single question (one exchange)
6% of a Google search.
6% of a Google search.
About 1.3× a drop of water and 1% of a teaspoon of water.
- Where it runsSelf-hosted / Edge (16GB)
- Grid carbon intensity458 g CO₂/kWh · 32% renewable
- Compared with a Google searchabout the same as a search (0.3 Wh)
- Water~14,925 queries to fill 1 litre
Where GPT-OSS 20B sits
#2 of 199 by energy per query. Lighter than 99% of all models. Lit dots are the same kind of model; click any dot to open it.
Want something lighter?
GPT-OSS 20B is already among the lighter well-known options for this kind of work, so there isn’t a clearly lighter alternative to suggest.
What a month of use adds up to
Pick how much you use it. Based on one quick use at a time, five days a week.
- Energy
- 1.9 Wh
- Carbon
- 0.83 g
- Water
- 7 mL
A month of 5 chats a day, 5 days a week. The energy is 16% of a phone charge.
Work out your own →- Energy
- 7.8 Wh
- Carbon
- 3.3 g
- Water
- 29 mL
A month of 20 chats a day, 5 days a week. The energy is 65% of a phone charge.
Work out your own →- Energy
- 39.0 Wh
- Carbon
- 16.7 g
- Water
- 145 mL
A month of 100 chats a day, 5 days a week. The energy is about 3.2× a phone charge and 35% of a kettle boil.
Work out your own →How does GPT-OSS 20B compare?
Next to similar models. Bars use a log scale, so each step is a big jump.
Detailed breakdown
Energy
We estimate GPT-OSS-20B at ~0.018 Wh for a quick single question, taken from the CLEER dataset (Sustainable AI Group, Sept 2026). The figure covers the full facility — accelerators, host servers, idle capacity and cooling. CLEER benchmarked the open weights directly (B200 x1 · batch 256). On the same basis a typical multi-turn chat comes to ~0.06 Wh and a typical agentic coding session to ~3.8 Wh, so how the model is used matters far more than which prompt you send. This replaces our earlier ~0.35 Wh estimate, which was extrapolated from 2025 benchmark figures on older hardware.
Power source & carbon
Fully open-weight. Can run on a laptop GPU, making it one of the most accessible capable models.
Water use
At ~0.067 mL per quick question, derived from the energy estimate at the ~3.7 L/kWh water factor we apply to this provider's infrastructure. Water scales directly with energy, so longer chats and agentic sessions use proportionally more.
About GPT-OSS 20B
At 0.02 Wh per query, GPT-OSS 20B is one of the most energy-efficient AI models in existence — ranked #2 across our entire dataset of 199 models. With just 20B parameters, it proves that useful AI doesn't have to be expensive to run. As an open-source model, it can be deployed on any hardware, including personal devices. OpenAI released it in August 7, 2025 as a text and chat model: compact open-weight model — runs on 16gb edge devices.
These figures are estimates derived from hardware specifications and API benchmarks — OpenAI has not published official energy data for GPT-OSS 20B. Actual consumption may vary significantly depending on batching, quantisation, and infrastructure optimisations that we cannot observe from outside.
GPT-OSS 20B in context
Key insights
- Ranked #2 most energy-efficient model in our entire dataset
- Uses less than a third of the average energy for text and chat models
- Open-source weights — can be self-hosted on infrastructure you control
The OpenAI GPT family
One chart per tier, on a shared time axis. Hover a dot for details; click to open that version.
Show 3 more tiers
32 versions across 8 tiers. Lightest: GPT-OSS 20B (0.02 Wh); heaviest: GPT Image 2 (5.0 Wh). Charts share a time axis but each tier has its own energy scale, so its trend is clear; the percentage is the change from a tier’s first version to its latest.
What does your GPT-OSS 20B usage cost the planet?
Use the calculator to estimate your personal footprint based on how often you use GPT-OSS 20B.
Frequently asked questions
How much energy does GPT-OSS 20B use per query?
Each GPT-OSS 20B query consumes approximately 0.02 Wh of energy. This is about the same as a traditional Google search (~0.3 Wh).
What is GPT-OSS 20B's carbon footprint?
Based on the carbon intensity of Self-hosted / Edge (16GB), each query produces approximately 0.01 g of CO2. The grid in this region has a carbon intensity of 458 g CO2/kWh with 32% renewable energy.
How much water does GPT-OSS 20B use?
Each query consumes approximately 0.07 mL of water, primarily used for cooling the data centers that process the request.
How does GPT-OSS 20B compare to a Google search?
A GPT-OSS 20B query uses about the same as a Google search in terms of energy. A Google search uses approximately 0.3 Wh, while GPT-OSS 20B uses 0.02 Wh.
Technical details
- Architecture
- Dense Transformer (open-weight, compact)
- Parameters
- 20B
- Context window
- 128,000 tokens
- Release date
- 2025-08-07
- Open source
- Yes
- Training data cutoff
- 2025-06