gpt-oss

ActiveOpen weightsLast updated 5 October 2026 · Robert Levandovsky

Verdict

Evidence: Editorial assessment

gpt-oss is OpenAI's pair of open-weight reasoning models, released under Apache 2.0, the most permissive licence in this group. The 20b model is the smaller, easier-to-host option; the 120b model trades hardware cost for capability. Choose gpt-oss when licence simplicity matters more than having the largest model.

Quick facts

gpt-oss
MakerOpenAI
CategoryAI Models & Open-Weight LLMs
Pricing modelOpen weights
StatusActive
Official sitegithub.com
Data last checked5 October 2026

Key facts and sources

Verified = read on the provider's own page. Reported = third-party source, not yet confirmed. Claimed by provider = vendor claim, not tested by aixjo.
ClaimDetailEvidenceSource
Model details and licenceOpenAI's open-weight models gpt-oss-120b (117B parameters, 5.1B active) and gpt-oss-20b (21B parameters, 3.6B active) are released under the Apache 2.0 licence.Verifiedopenai/gpt-oss on GitHub, 5 October 2026

Pricing

No prices are shown because we have not confirmed them on gpt-oss's own pricing page. Go to gpt-oss

Who should use it

  • Commercial products that need a permissive licence
  • Local or on-premise deployment of a mid-sized reasoning model

Who should skip it

  • You need the most capable open model regardless of size (compare Qwen3.8 or Kimi K3)

Limitations and risks

  • Smaller than 2026's largest open models.

How it compares: gpt-oss alternatives

Rating

Not yet rated. We show a rating only when at least four of six criteria have documented evidence. How we rate AI tools

Common questions

What licence does gpt-oss use?

Apache 2.0, according to OpenAI's gpt-oss repository.

What sizes does gpt-oss come in?

gpt-oss-120b (117B parameters, 5.1B active) and gpt-oss-20b (21B parameters, 3.6B active).

Sources