Singularity Logs

THE MODEL, IN DETAIL / openai

gpt-oss-120b

openai/gpt-oss-120bOpen weightsReasoning

gpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and general-purpose production use cases. It activates 5.1B parameters per forward pass and is optimized...

OpenRouter
Source-linked model intelligenceCatalog added 5 Aug 2025Specifications observed 17 Oct 2026

GPQA Diamond

67.1%

Source-reported

Context window

131.1K

65.5K max output tokens

Input price

$0.15

USD / million input tokens

Output price

$0.6

USD / million output tokens

Inference providers

22

24 provider routes observed

Measured output speed

—

No model-level measurement

01 / PERFORMANCE

A clearer view of capability.

Read the score, understand the context, and explore where this model sits among its peers.

Performance fingerprint

Individual published accuracy scores · not a composite index

At least three published accuracy scores are needed for this view.

Performance meets price

GPQA Diamond · same published source and version

Higher score · lower cost20$037.5$0.201255$0.402572.5$0.603890$0.805Qwen3.5-397B-A17B GPQA Diamond: 88.4 Input: $0.5/M tokensKimi-K2.5 GPQA Diamond: 87.6 Input: $0.5/M tokensQwen3.5-122B-A10B GPQA Diamond: 86.6 Input: $0.26/M tokensGLM-4.7-Flash GPQA Diamond: 75.2 Input: $0.0605/M tokensDeepSeek-R1 GPQA Diamond: 71.5 Input: $0.7/M tokensgpt-oss-120b GPQA Diamond: 67.1 Input: $0.15/M tokensgpt-oss-120bgpt-oss-20b GPQA Diamond: 56.8 Input: $0.018/M tokensLlama-3.1-8B-Instruct GPQA Diamond: 30.4 Input: $0.05/M tokensInput price · USD / million tokens
Qwenmoonshotaizai-orgdeepseek-aiopenaimeta-llama

In good company

GPQA Diamond · OpenEvals / Hugging Face

88.4
Qwen
87.6
moonshotai
86.6
Qwen
79.2
nvidia
79.1
LGAI-EXAONE
76.2
Qwen
75.2
zai-org
74.4
jdopensource
71.5
deepseek-ai
67.1
This model

Rank #21 in this published cohort. 27 models in the source cohort. Showing selected nearby models. Evaluation settings can differ.

Evidence, by task.

Coverage is not a capability score.

2published results
1 evaluation source
Reasoning2 evaluations
CodingAwaiting evidence
MathematicsAwaiting evidence
VisionAwaiting evidence
AgenticAwaiting evidence
Long contextAwaiting evidence

Documented features, task evaluations, and preference rankings describe different things. Missing coverage does not imply a model cannot perform a task.

Explore our methodology

02 / BENCHMARK EXPLORER

The numbers behind the model.

Original results, exact versions, and a direct path to the evidence behind every score.

2 published results
020406080100GPQA Diamond: 67.1% · Published snapshot 2026-03-16 · OpenEvals / Hugging Face67.1%GPQA DiamondHigher is betterHumanity’s Last Exam: 5.2% · Published snapshot 2026-03-16 · OpenEvals / Hugging Face5.2%Humanity’s Last …Higher is better

2 percentage-based evaluations. Each benchmark measures a different task; compare results within the same evaluation and protocol.

Published benchmark results for gpt-oss-120b
BenchmarkScoreVersion & metricSourcePublishedEvidence
Published snapshot 2026-03-16Accuracy · Higher is betterOpenEvals / Hugging FaceSource-reported15 Mar 2026
Published snapshot 2026-03-16Accuracy · Higher is betterOpenEvals / Hugging FaceSource-reported15 Mar 2026

Artificial Analysis indices

Published in the OpenRouter catalog · source-defined index values

Coding Index30.4
Agentic Index3.7
Intelligence Index11.6

Catalog observation: 17 Oct 2026. Index versions and evaluation settings are not supplied in this feed. These are not Singularity composite scores.

Artificial Analysis

Design Arena

Source-reported Elo across individual design categories

3d902
models arena · Rank #118 · 29.4% win rate
Codecategories965
models arena · Rank #128 · 33.4% win rate
Dataviz987
models arena · Rank #116 · 43.6% win rate
Gamedev995
models arena · Rank #118 · 40.5% win rate
Uicomponent926
models arena · Rank #121 · 35.7% win rate
Website969
models arena · Rank #131 · 32.5% win rate

Catalog observation: 17 Oct 2026. Category ratings have separate comparison pools.

03 / COST & VALUE

Know what your workload costs.

Explore token prices, see where the spend goes, and model a workload of your own.

Input$0.15per million tokens
Output$0.6per million tokens
Cache read$0.075per million tokens
Cache write—per million tokens

Workload cost calculator

Change the numbers. See the trade-offs.

Token estimate only. Excludes cache writes, tools, and other provider fees. Actual token usage and provider limits vary.

Your estimated spend

1,000 requests · 11M tokens

$2.1estimated total
Uncached input$1.5
Cached input$0
Output$0.6
Per request$0.0021
Without cache discount$2.1
Estimated cache savings$0

04 / INFERENCE NETWORK

One model. Different ways to run it.

Compare inference routes by price, context, quantization, and reported availability.

22 providers · 24 routes
Provider routes for gpt-oss-120b, observed 17 Oct 2026
Provider / routeInput / MOutput / MCache / MContextOutput limitSpeedLatencyAvailability · 1dRates
NNear AIFP4$0.03$0.15$0.015131.1K118K——99.86%
VVeniceINT4$0.03$0.15$0.03128K32.8K——99.93%
CCoreWeaveFP4$0.03$0.17$0.03131.1K118K——99.51%
DDekaLLMBF16$0.03$0.18$0.03131.1K118K——99.93%
AAkashMLBF16$0.03$0.187$0.03131.1K118K——100%
DDeepInfraBF16$0.037$0.17—131.1K118K——97.78%
CCrusoeBF16$0.05$0.25$0.05131.1K118K——100%
MMancer 2FP8$0.05$0.25—131.1K118K——99.72%
NNovitaFP4$0.05$0.25—131.1K32.8K——98.33%
DDigitalOceanQuantization not disclosed$0.06$0.42$0.012128K4.1K——100%
GGoogleQuantization not disclosed$0.09$0.36—131.1K118K——8.9%
BBaseTenFP4$0.1$0.5$0.1128.1K115.3K——100%
BBaseTenFP4$0.1$0.5$0.1128.1K115.3K——99.65%
PParasailFP4$0.1$0.75$0.055131.1K118K——100%
SSambaNovaQuantization not disclosed$0.14$0.95—131.1K118K——99.85%
AAmazon BedrockQuantization not disclosed$0.15$0.6—131.1K118K——99.85%
NNebiusFP4$0.15$0.6—131.1K118K——100%
AAmazon BedrockQuantization not disclosed$0.15$0.6—131.1K118K——98.77%
SSiliconFlowFP8$0.15$0.6$0.075131.1K8.2K——78.37%
PPhalaQuantization not disclosed$0.15$0.6—131.1K32.8K——99.69%
TTogetherQuantization not disclosed$0.15$0.6—131.1K118K——84.23%
GGroqQuantization not disclosed$0.15$0.6$0.075131.1K65.5K——100%
MMaraQuantization not disclosed$0.15$0.75—131.1K118K——97.11%
CCerebrasFP16$0.35$0.75$0.35131.1K41K——100%

Snapshot observed 17 Oct 2026. Speed and latency are the reported 30-minute median when available. This is a dated snapshot, not live monitoring.

Provider price comparison

USD per million tokens · input and output rates

InputOutput
Near AIfp4
Input
$0.03
Output
$0.15
Veniceint4
Input
$0.03
Output
$0.15
CoreWeavefp4
Input
$0.03
Output
$0.17
DekaLLMbf16
Input
$0.03
Output
$0.18
AkashMLbf16
Input
$0.03
Output
$0.187
DeepInfrabf16
Input
$0.037
Output
$0.17
Crusoebf16
Input
$0.05
Output
$0.25
Mancer 2fp8
Input
$0.05
Output
$0.25
Novitafp4
Input
$0.05
Output
$0.25
DigitalOcean
Input
$0.06
Output
$0.42

Reported availability

Provider availability over the selected observation window

Near AI · FP499.86%
Venice · INT499.93%
CoreWeave · FP499.51%
DekaLLM · BF1699.93%
AkashML · BF16100%
DeepInfra · BF1697.78%
Crusoe · BF16100%
Mancer 2 · FP899.72%
Novita · FP498.33%
DigitalOcean100%

Output speed

Reported 30-minute p50 · tokens per second

No provider throughput measurements were returned in this snapshot.

Response latency

Reported 30-minute p50 · seconds

No provider latency measurements were returned in this snapshot.

Context by provider

Actual route limits can differ from the catalog maximum

Near AI · FP4131.1K
Venice · INT4128K
CoreWeave · FP4131.1K
DekaLLM · BF16131.1K
AkashML · BF16131.1K
DeepInfra · BF16131.1K
Crusoe · BF16131.1K
Mancer 2 · FP8131.1K
Novita · FP4131.1K
DigitalOcean128K

05 / CAPABILITIES & MODALITIES

What goes in. What comes out.

A map of documented interfaces and API features, separate from benchmark performance.

INPUT MODALITIES

Text
gpt-oss-120bReasoning supported

OUTPUT MODALITIES

Text

Feature support

Listed catalog API parameters and documented pricing features

ReasoningSupported
Tool callingSupported
Tool choiceSupported
Structured outputsSupported
Response formatSupported
Reasoning detailsSupported
Seed controlSupported
Stop sequencesSupported
Prompt cachingSupported
TemperatureSupported

Room to work

Context capacity and completion limits

131.1Ktotal context tokens
Context window
65.5Kmax output

Output tokens share the total context budget. The maximum completion length is a separate limit, and individual provider routes may have smaller limits.

Catalog context
131,072 tokens
Maximum output
65,536 tokens
Top-provider context
131,072 tokens

Features vary by route

Check the provider before depending on a specific API feature.

API feature support by provider
Provider routeReasoningToolsStructured outputResponse formatImplicit caching
Near AI · FP4SupportedSupportedSupportedSupportedSupported
Venice · INT4SupportedSupportedNot listedNot listedSupported
CoreWeave · FP4SupportedSupportedSupportedSupportedSupported
DekaLLM · BF16SupportedSupportedSupportedSupportedSupported
AkashML · BF16SupportedSupportedSupportedSupportedSupported
DeepInfra · BF16SupportedSupportedSupportedSupportedNot listed
Crusoe · BF16SupportedSupportedSupportedSupportedSupported
Mancer 2 · FP8SupportedSupportedSupportedSupportedNot listed
Novita · FP4SupportedSupportedNot listedSupportedNot listed
DigitalOceanSupportedNot listedNot listedNot listedSupported
GoogleSupportedNot listedSupportedSupportedNot listed
BaseTen · FP4SupportedSupportedSupportedSupportedSupported
BaseTen · FP4SupportedSupportedSupportedSupportedSupported
Parasail · FP4SupportedSupportedSupportedSupportedSupported
SambaNovaSupportedSupportedNot listedNot listedNot listed
Amazon BedrockSupportedNot listedNot listedNot listedNot listed
Nebius · FP4SupportedSupportedSupportedSupportedNot listed
Amazon BedrockSupportedNot listedNot listedNot listedNot listed
SiliconFlow · FP8SupportedNot listedSupportedSupportedSupported
PhalaSupportedSupportedSupportedSupportedNot listed
TogetherSupportedSupportedSupportedSupportedNot listed
GroqSupportedSupportedSupportedSupportedSupported
MaraSupportedSupportedSupportedSupportedNot listed
Cerebras · FP16SupportedSupportedSupportedSupportedSupported

06 / UNDER THE HOOD

The complete specification sheet.

Exact identities, model architecture, limits, and the defaults that shape a request.

Identity & architecture

Developer
openai
Exact identity
openai/gpt-oss-120b
Subject
Model
Access
Open weights
License
Not disclosed
Parameters
120.4 billion
Tokenizer
GPT
Instruction format
Not disclosed

Reasoning & controls

Reasoning
Supported
Mandatory reasoning
Yes
Reasoning by default
Not documented
Published configuration
Not disclosed
Supported reasoning efforts
high, medium, low
Default reasoning effort
medium
API parameters
21
Provider moderation
No
Quantization variants
fp4, int4, bf16, fp8, fp16

Dates & limits

Catalog added
5 Aug 2025
Knowledge cutoff
2024-06-30
Catalog expiration
Not documented
Specifications observed
17 Oct 2026
Providers observed
17 Oct 2026
Request limits
Provider-specific / not supplied
Canonical catalog ID
openai/gpt-oss-120b

API parameters & defaults

Availability is catalog-wide. Defaults can vary across providers.

frequency_penaltySupported · default not supplied
include_reasoningSupported · default not supplied
logit_biasSupported · default not supplied
logprobsSupported · default not supplied
max_tokensSupported · default not supplied
min_pSupported · default not supplied
presence_penaltySupported · default not supplied
reasoningSupported · default not supplied
reasoning_effortSupported · default not supplied
repetition_penaltySupported · default not supplied
response_formatSupported · default not supplied
seedSupported · default not supplied
stopSupported · default not supplied
structured_outputsSupported · default not supplied
temperatureSupported · default not supplied
tool_choiceSupported · default not supplied
toolsSupported · default not supplied
top_aSupported · default not supplied
top_kSupported · default not supplied
top_logprobsSupported · default not supplied
top_pSupported · default not supplied

07 / BUILD WITH THIS MODEL

From exploration to your first request.

Use the documented catalog identifier and keep your API key in an environment variable.

Make an API request

const response = await fetch(
  "https://openrouter.ai/api/v1/chat/completions",
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.OPENROUTER_API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      "model": "openai/gpt-oss-120b",
      "messages": [
        {
          "role": "user",
          "content": "Explain the trade-offs in this design."
        }
      ]
    }),
  },
);
if (!response.ok) throw new Error(`HTTP ${response.status}`);
const data = await response.json();
console.log(data.choices[0].message.content);

08 / PROVENANCE & HISTORY

Every number has a story.

Follow the source, check the observation date, and distinguish published results from model specifications.

The observation trail

Recorded events, in chronological order

Added to the model catalog

Catalog creation date, as recorded by OpenRouter.

GPQA Diamond result published

67.1% · OpenEvals / Hugging Face

Humanity’s Last Exam result published

5.2% · OpenEvals / Hugging Face

Model specifications observed

Catalog pricing, limits, and supported parameters captured.

Provider routes observed

24 inference routes in this snapshot.

Observation dates describe when data was captured. They do not represent newly run evaluations or continuous performance monitoring.

Published observation ledger

Historical evidence, with source and version preserved.

Published observation history for gpt-oss-120b
ObservedBenchmarkScoreVersionSourceVerification
13 Oct 2026Humanity’s Last ExamPublished snapshot 2026-03-16OpenEvals / Hugging FaceSource-reported
13 Oct 2026GPQA DiamondPublished snapshot 2026-03-16OpenEvals / Hugging FaceSource-reported

09 / A FEW MORE DETAILS

Questions worth asking.

Quick answers, drawn from this model’s connected specifications.

How much does gpt-oss-120b cost?

Catalog starting rates are $0.15 per million input tokens and $0.6 per million output tokens. Cache-read pricing is $0.075 per million tokens. Compare provider routes and service tiers above for their specific prices.

How much context can it handle?

131,072 tokens in the catalog context window. The catalog lists a maximum of 65,536 completion tokens. Provider limits can differ.

Does it support tools and structured output?

Tool calling: listed as supported. Structured outputs: listed as supported. See the provider feature matrix for route-specific differences.

How should I interpret these benchmark results?

Read the benchmark name, exact version, metric, source, and model configuration together. Different evaluation settings can change the result. Preference ratings and source-defined composite indices remain separate from task accuracy.

Are these prices and provider measurements live?

This page uses dated source snapshots. Catalog data was observed on 17 Oct 2026. Provider routes were observed on 17 Oct 2026. Missing measurements are not estimated or treated as zero.

KEEP EXPLORING

More from openai.

Explore related identities from the same developer, or take them into a side-by-side comparison.

THE RIGHT MODEL STARTS WITH THE RIGHT CONTEXT.

Put the details side by side.

Explore the trade-offs in a direct model comparison.

Compare gpt-oss-120b
Source-backed evidence. Different tasks. A clearer picture.
AboutMethodologyBenchmark dashboard