OpenAI: gpt-oss-20b
Analysis Summary
OpenAI's gpt-oss-20b, released in August 2025, is a compact open-weights model that punches well above its size class. Its intelligence index of 24.5 and coding index of 18.5 are strong, but the standout figures are its agentic index of 35.4, math index of 89.3, and AIME score of 0.893, which are near-frontier for a model at this price point. LiveCodeBench at 0.777 and IFBench at 0.651 confirm excellent coding and instruction-following capability. Tool use and function calling are supported.
For businesses, gpt-oss-20b is a strong fit for autonomous coding agents, mathematical analysis pipelines, and structured automation workflows where cost efficiency matters. Its agentic performance is particularly notable, The 131K context window is adequate for most business document tasks. No vision support is the main capability gap.
At $0.029 input and $0.14 output per million tokens, it is one of the best value propositions for agentic and coding workloads from a major provider. Teams building cost-efficient coding agents or high-volume automation pipelines will find it a compelling choice, backed by OpenAI's infrastructure and reliability.
Assessed June 9, 2026
Editorial notes
OpenAI's gpt-oss-20b delivers world-class agentic performance, strong coding benchmarks, and exceptional math reasoning at very low cost from a trusted provider.
Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?
Performance Profile
How OpenAI: gpt-oss-20b compares
OpenAI: gpt-oss-20b ranks #168 of 378 AI models we track for overall intelligence, #188 of 315 for coding, #147 of 289 for agentic tasks. Its 131K-token context window is larger than 60% of the models we list. At $0.03 per million input tokens it is cheaper than 76% of comparable models.
About OpenAI: gpt-oss-20b
gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for..
Capabilities
Performance Indices
Source: Artificial Analysis
Benchmark Scores
Intelligence
Technical
Content
Benchmark data from Artificial Analysis and Hugging Face
How does OpenAI: gpt-oss-20b stack up?
Compare side-by-side with other professional models.
Model Information
| OpenRouter ID |
openai/gpt-oss-20b
|
| Provider | openai |
| Release Date | August 5, 2025 |
| Context Length | 131,072 tokens |
| Status | Active |
Pricing
| Token Type | Cost per 1M tokens | Cost per 1K tokens |
|---|---|---|
| Input | $0.03 | $0.000029 |
| Output | $0.14 | $0.000140 |
Live Performance
Live endpoint metrics, refreshed every 30 minutes.
Leaderboard Categories
External Resources
Frequently asked questions about OpenAI: gpt-oss-20b
How much does OpenAI: gpt-oss-20b cost?
OpenAI: gpt-oss-20b costs $0.03 per million input tokens and $0.14 per million output tokens.
What is the context window of OpenAI: gpt-oss-20b?
OpenAI: gpt-oss-20b has a context window of 131,072 tokens (131K).
Is OpenAI: gpt-oss-20b good for coding?
On our coding benchmark index, OpenAI: gpt-oss-20b ranks #188 of 315 models, placing it in the broader range of the field for code generation and debugging.
What can OpenAI: gpt-oss-20b do?
OpenAI: gpt-oss-20b supports tool use and function calling.
Who created OpenAI: gpt-oss-20b?
OpenAI: gpt-oss-20b is developed by OpenAI and was released on August 5, 2025.
