Skip to content
  • Models
  • Rankings
  • Ori
OpenRouterOpenRouter
© 2026 OpenRouter, Inc

Product

  • Chat
  • Rankings
  • Benchmarks
  • Apps
  • Discover
  • Models
  • Collections
  • Providers
  • Tools
  • Pricing
  • Business
  • Enterprise
  • Labs

Company

  • About
  • Blog
  • Careers
    Hiring
  • Privacy
  • Terms of Service
  • Trust Center
  • Support
  • Works With OR
  • Data
  • Brand

Developer

  • Documentation
  • API Reference
  • Developer Platform
  • Status
  • AI Site Map

Connect

  • Discord
  • GitHub
  • LinkedIn
  • X
  • YouTube
Favicon for openai

OpenAI: gpt-oss-20b

openai/gpt-oss-20b

Model weights
Compare

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 lower-latency inference and deployability on consumer or single-GPU hardware. The model is trained in OpenAI’s Harmony response format and supports reasoning level configuration, fine-tuning, and agentic capabilities including function calling, tool use, and structured outputs.

Modalities
In / Out Price
$0.018 / $0.09per 1M
Context
131K
Released
Aug 5, 2025
Knowledge Cutoff
Jun 2024
Compare
ProvidersPricingPerformanceUptimeBenchmarksAppsActivityFAQExplore

Providers

Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick — Balanced (price + speed), Nitro (fastest), Floor (cheapest), or Exacto (highest tool-calling accuracy).

Pricing

The average price customers actually pay for this model, next to the prices providers post. Caching and discounts mean the price actually paid is often well below the listed one.

Performance

Throughput is how fast the model writes (tokens per second — higher is better). Latency is total round-trip time (lower is better). TTFT is time-to-first-token — how long before you see anything appear (lower is better).

Uptime

Uptime is the percentage of the past 3 days that at least one provider was responding to requests. Availability is the percentage of time that inference was successfully served. OpenRouter continuously monitors and uses the next-best provider when one returns an error.

Benchmarks

Scores on standardized evaluations and this model’s own outputs on shared benchmark prompts. Higher percentages are better — and each rank shows where this model lands among the models evaluated on that source.

Benchmark score summary for OpenAI: gpt-oss-20b (Artificial Analysis and Design Arena)
SourceBenchmarkScore
Artificial Analysisgpt-oss-20b (low) GPQA Diamond61.1%
Artificial Analysisgpt-oss-20b (low) HLE5.3%
Artificial Analysisgpt-oss-20b (low) IFBench57.8%
Artificial Analysisgpt-oss-20b (low) τ²-Bench Telecom50.3%
Artificial Analysisgpt-oss-20b (low) AA-LCR31.0%
Artificial Analysisgpt-oss-20b (low) CritPt0.0%
Artificial Analysisgpt-oss-20b (low) Terminal-Bench Hard4.5%
Artificial Analysisgpt-oss-20b (low) AA-Omniscience Accuracy15.1%
Artificial Analysisgpt-oss-20b (low) AA-Omniscience Non-Hallucination Rate13.3%
Artificial Analysisgpt-oss-20b (high) Intelligence Index9.0
Artificial Analysisgpt-oss-20b (high) Coding Index20.7
Artificial Analysisgpt-oss-20b (high) Agentic Index1.2
Artificial Analysisgpt-oss-20b (high) GPQA Diamond68.8%
Artificial Analysisgpt-oss-20b (high) HLE11.0%
Artificial Analysisgpt-oss-20b (high) IFBench65.1%
Artificial Analysisgpt-oss-20b (high) τ²-Bench Telecom60.2%
Artificial Analysisgpt-oss-20b (high) AA-LCR34.7%
Artificial Analysisgpt-oss-20b (high) GDPval-AA0.0%
Artificial Analysisgpt-oss-20b (high) CritPt1.4%
Artificial Analysisgpt-oss-20b (high) SciCode38.9%
Artificial Analysisgpt-oss-20b (high) Terminal-Bench Hard10.6%
Artificial Analysisgpt-oss-20b (high) AA-Omniscience Accuracy16.0%
Artificial Analysisgpt-oss-20b (high) AA-Omniscience Non-Hallucination Rate5.9%
Design ArenaGPT OSS 20B Models Arena Data Visualization Elo940
Design ArenaGPT OSS 20B Models Arena Website Elo859

Apps

Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for.

Activity

Token volume and request traffic to this model over time.

API

Drop-in code to call this model. OpenRouter's API is OpenAI-compatible — most SDKs work by just swapping the base URL. The only thing that changes between models is the model slug below.

Explore more models

AI Model RankingsRanking
$0.018$0.09$0.00902.83s30 tps
100.00%
$0.020$0.10--0.76s50 tps
99.95%
$0.029$0.14$0.02900.72s34 tps
99.84%
$0.030$0.13$0.03000.32s121 tps
100.00%
$0.030$0.14--0.23s84 tps
99.79%
$0.030$0.15$0.02000.66s92 tps
99.58%
$0.040$0.15--0.77s64 tps
99.98%
$0.070$0.15--0.52s264 tps
100.00%
$0.070$0.25--0.33s219 tps
95.50%
$0.070$0.30--0.57s149 tps
99.82%
$0.075$0.30$0.03750.31s432 tps
99.60%
$0.040$0.18--1.03s45 tps
91.99%
Not used in Standard routing:Why these endpoints are not used
$0.070$0.15--0.43s143 tps
99.06%
Throughput
432tok/s

P50, best across providers

Latency
0.23s

P50, best provider

AutoExacto benchmarks
GPQA DiamondTAU-BenchDeepInfra68.4%--AkashML66.0%--SiliconFlow64.8%--auto-routing66.5%54.7%NovitaAI58.1%--
+6 more providers
Uptime (3d)The model was reachable. Request routed to a provider.
100.00%
Availability (3d)The model returned inference from any provider. Errors and empty responses count against it.
95.57%

Availability over the last 3 days

Last 72 hours
Availability 95.57%
3 Days Ago2 Days AgoYesterdayNow

Availability over the last 24 hours

OpenRouter Availability
94.44%
Without Routing
89.39%

When an error occurs in an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it. You can access per-provider uptime data programmatically through the Endpoints API. Learn more about our load balancing and customization options.

1.
Favicon for https://heyditto.ai/
HeyDitto
new
163Btokens
2.
Favicon for https://jobleads.com/
JobLeads LLM
Use AI to find jobs for you
28.6Btokens
3.
Favicon for https://centralcommand.io/
Central Command
new
27.9Btokens
4.
Favicon for https://craftrpgs.com/?utm_source=openrouter
Craft
Build and play AI RPGs
14.5Btokens
5.
Favicon for https://fvchat.app/
FVChat
new
14.2Btokens

Frequently asked questions

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 lower-latency inference and deployability on consumer or single-GPU hardware.

gpt-oss-20b costs $0.018/M input tokens and $0.09/M output tokens, with separate rates for Cache Read at $0.009/M tokens.

gpt-oss-20b has a 131,072 token context window. It supports up to 32,768 completion tokens.

Yes. gpt-oss-20b accepts tools and tool_choice for function calling on 9 of the 12 providers serving it, and requests that send tools are routed to those providers. It also supports structured outputs via a JSON schema in response_format.

gpt-oss-20b is served by 12 providers on OpenRouter: Darkbloom, AkashML, DekaLLM, CoreWeave, DeepInfra, Parasail, NovitaAI, SiliconFlow and 4 more. Requests are routed to the best available provider, with automatic failover to the others, and you can pin or exclude providers with provider routing.

gpt-oss-20b was released on August 5, 2025. Its knowledge cutoff is June 30, 2024.

More models from OpenAI

GPT-6.1 Sol Pro

GPT-6.1 Sol Pro is the same underlying model as GPT-6.1 Sol, served with reasoning.mode set to pro for higher-quality responses on complex tasks.

Cost note: pro mode spends far more reasoning tokens per request, so a typical request costs several times more than the same request on GPT-6.1 Sol and takes much longer to complete. It is intended for hard, high-stakes problems where the extra accuracy justifies the cost. For everyday coding, agentic, and chat workloads, use GPT-6.1 Sol instead.

Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode

Text1.1M context$2 / $10
GPT-6.1 Sol

GPT-6.1 Sol is an upgrade to GPT-6 Sol from OpenAI, positioned below the flagship GPT-6 Astra in the GPT-6 series. It is suited for agentic coding, computer use, document-heavy professional work, and multi-step business workflow automation, and approaches Astra-level results on these tasks at a much lower cost. Compared with GPT-6 Sol, it makes fewer factual errors and is more reliable at respecting explicit restrictions and user intent during agentic tasks.

Text1.1M context$2 / $10
GPT-6 Luna Pro

GPT-6 Luna Pro is the same underlying model as GPT-6 Luna, served with reasoning.mode set to pro for higher-quality responses on complex tasks.

Cost note: pro mode spends far more reasoning tokens per request, so a typical request costs several times more than the same request on GPT-6 Luna and takes much longer to complete. It is intended for hard, high-stakes problems where the extra accuracy justifies the cost. For everyday coding, agentic, and chat workloads, use GPT-6 Luna instead.

Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode

Text1.1M context$0.10 / $0.50
GPT-6 Luna Pro

GPT-6 Luna Pro is the same underlying model as GPT-6 Luna, served with reasoning.mode set to pro for higher-quality responses on complex tasks.

Cost note: pro mode spends far more reasoning tokens per request, so a typical request costs several times more than the same request on GPT-6 Luna and takes much longer to complete. It is intended for hard, high-stakes problems where the extra accuracy justifies the cost. For everyday coding, agentic, and chat workloads, use GPT-6 Luna instead.

Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode

Text1.1M context$0.05 / $0.25
GPT-6 Luna

GPT-6 Luna is the fast, cost-efficient model in OpenAI's GPT-6 series, positioned below GPT-6 Sol. It is suited for high-volume and latency-sensitive workloads such as chat, classification, and lightweight agentic tasks, and at higher reasoning effort it can take on complex software engineering and computer-use tasks that previously called for a Sol-tier model. It shares the GPT-6 family's gains in factual reliability and its clearer, more concise communication style.

Text1.1M context$0.10 / $0.50
GPT-6 Luna

GPT-6 Luna is the fast, cost-efficient model in OpenAI's GPT-6 series, positioned below GPT-6 Sol. It is suited for high-volume and latency-sensitive workloads such as chat, classification, and lightweight agentic tasks, and at higher reasoning effort it can take on complex software engineering and computer-use tasks that previously called for a Sol-tier model. It shares the GPT-6 family's gains in factual reliability and its clearer, more concise communication style.

Text1.1M context$0.05 / $0.25
GPT-6 Sol Pro

GPT-6 Sol Pro is the same underlying model as GPT-6 Sol, served with reasoning.mode set to pro for higher-quality responses on complex tasks.

Cost note: pro mode spends far more reasoning tokens per request, so a typical request costs several times more than the same request on GPT-6 Sol and takes much longer to complete. It is intended for hard, high-stakes problems where the extra accuracy justifies the cost. For everyday coding, agentic, and chat workloads, use GPT-6 Sol instead.

Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode

Text1.1M context$2 / $10
GPT-6 Sol Pro

GPT-6 Sol Pro is the same underlying model as GPT-6 Sol, served with reasoning.mode set to pro for higher-quality responses on complex tasks.

Cost note: pro mode spends far more reasoning tokens per request, so a typical request costs several times more than the same request on GPT-6 Sol and takes much longer to complete. It is intended for hard, high-stakes problems where the extra accuracy justifies the cost. For everyday coding, agentic, and chat workloads, use GPT-6 Sol instead.

Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode

Text1.1M context$1 / $5
GPT-6 Sol

GPT-6 Sol is the cost-efficient high-end model in OpenAI's GPT-6 series, positioned below the flagship GPT-6 Astra and above the fast GPT-6 Luna tier. It is suited for demanding professional work, agentic coding, business workflow automation, and computer use, and is particularly strong at long-horizon software engineering tasks in real codebases. It approaches Astra-level factual reliability at a much lower cost and shares Astra's clearer, more concise communication style in technical and coding conversations.

Text1.1M context$2 / $10
GPT-6 Sol

GPT-6 Sol is the cost-efficient high-end model in OpenAI's GPT-6 series, positioned below the flagship GPT-6 Astra and above the fast GPT-6 Luna tier. It is suited for demanding professional work, agentic coding, business workflow automation, and computer use, and is particularly strong at long-horizon software engineering tasks in real codebases. It approaches Astra-level factual reliability at a much lower cost and shares Astra's clearer, more concise communication style in technical and coding conversations.

Text1.1M context$1 / $5
GPT Astra Latest

This model always redirects to the latest model in the GPT Astra family.

Text1.1M context
GPT Sol Latest

This model always redirects to the latest model in the GPT Sol family.

Text1.1M context
GPT Terra Latest

This model always redirects to the latest model in the GPT Terra family.

Text1.1M context
GPT Luna Latest

This model always redirects to the latest model in the GPT Luna family.

Text1.1M context
GPT Image 2.5 Sunburst

GPT Image 2.5 Sunburst is an image generation and editing model from OpenAI, positioned as the precision-oriented tier of the GPT Image 2.5 series. It is suited to detailed creative work where editing accuracy matters more than generation speed, via the dedicated Images API.

Image$8 / $30
GPT Image 2.5 Flare

GPT Image 2.5 Flare is an image generation and editing model from OpenAI, positioned as the speed-oriented tier of the GPT Image 2.5 series. It is suited to high-volume everyday generation, creator content, and rapid prototyping via the dedicated Images API.

Image$8 / $30
GPT-6 Astra

GPT-6 Astra is OpenAI's flagship model for demanding end-to-end work. It is suited for advanced analysis, software engineering, deep research, scientific work, and document creation, with particular strengths in long-horizon agentic tasks that involve computer and browser use.

Text1.1M context$10 / $50
GPT-6 Astra

GPT-6 Astra is OpenAI's flagship model for demanding end-to-end work. It is suited for advanced analysis, software engineering, deep research, scientific work, and document creation, with particular strengths in long-horizon agentic tasks that involve computer and browser use.

Text1.1M context$5 / $25
GPT-6 Astra Pro

GPT-6 Astra Pro is the same underlying model as GPT-6 Astra, served with reasoning.mode set to pro for higher-quality responses on complex tasks.

Cost note: pro mode spends far more reasoning tokens per request, so a typical request costs several times more than the same request on GPT-6 Astra and takes much longer to complete. It is intended for hard, high-stakes problems where the extra accuracy justifies the cost. For everyday coding, agentic, and chat workloads, use GPT-6 Astra instead.

Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode

Text1.1M context$10 / $50
GPT-6 Astra Pro

GPT-6 Astra Pro is the same underlying model as GPT-6 Astra, served with reasoning.mode set to pro for higher-quality responses on complex tasks.

Cost note: pro mode spends far more reasoning tokens per request, so a typical request costs several times more than the same request on GPT-6 Astra and takes much longer to complete. It is intended for hard, high-stakes problems where the extra accuracy justifies the cost. For everyday coding, agentic, and chat workloads, use GPT-6 Astra instead.

Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode

Text1.1M context$5 / $25