Compare AI models

Specifications
Parameters
—21B
Context window
1M128k
API pricing
Cheapest input
—$0.018AkashML
Cheapest output
—$0.09AkashML
—$8.10Darkbloom · fp8
Benchmarks
Humanity's Last Exam · no tools
40.2%10.9%
Overview
CompanyGoogleOpenAI
Release dateMay 19 2026Aug 5 2025
AccessClosedOpen Weight
Model detailsView modelView model

Frequently asked questions

Gemini 3.5 Flash leads gpt-oss-20b on 1 of the 1 benchmark they both report (Humanity's Last Exam). gpt-oss-20b shipped 287 days before Gemini 3.5 Flash, so benchmark comparisons should account for the intervening progress.

Context windows are 1M (Gemini 3.5 Flash) vs 128k (gpt-oss-20b). Gemini 3.5 Flash is closed, while gpt-oss-20b is open weight.

On Humanity's Last Exam · no tools, Gemini 3.5 Flash leads at 40.2% vs gpt-oss-20b at 10.9%.