Inference
EN
OpenAI

GPT-OSS 120B

GPT-OSS 120B is a standard model from OpenAI in the GPT-OSS family. 25 benchmarks count toward its index, in 7 categories.

  • Open weights
  • Standard
  • Input: text
  • Released 2025-08
RankedEnough results for an index.
IndexAll categories, out of 100.43.4±4.2Median 51.0
CoverageIndex weight that has results.90%
SpeedOutput tokens per second.34/sMedian 67/s
First tokenSeconds to the first output token.0.3 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.15Median $0.60
Output / 1MUS dollars per 1M output tokens.$0.60Median $2.55
ContextMaximum tokens in one request.131KMedian 262K
EloLMArena rating and rank.1365±4 #197Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
openai/gpt-oss-120b
Max outputMaximum output tokens per request.
66K
Knowledge cutoffLast date of training data.
30 Jun 2024
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.075
WeightsLink to the published weights.
Hugging Face (opens in a new tab) ↗
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
26 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
CoreWeaveHost$0.03$0.17131K
Deep InfraHost$0.037$0.17131K
CrusoeHost$0.05$0.25131K
Novita AIHost$0.05$0.25131K
Google Vertex AICloud$0.09$0.36131K
SiliconFlowHost$0.05$0.45131K
OVHcloudCloud$0.09$0.47131K
BasetenHost$0.10$0.50128K
SambaNovaHost$0.22$0.59131K
Amazon BedrockCloud$0.15$0.60131K
DatabricksCloud$0.15$0.60131K
GroqHost$0.15$0.60131K
NebiusHost$0.15$0.60131K
OpenRouterRouter$0.15$0.60131K
Oracle CloudCloud$0.15$0.60128K
TinfoilHost$0.15$0.60131K
Together AIHost$0.15$0.60131K
IBM watsonxCloud$0.159$0.636131K
ScalewayCloud$0.168$0.673128K
DigitalOceanCloud$0.10$0.70128K
IO.NETHost$0.188$0.70131K
CerebrasHost$0.35$0.75131K
Cloudflare Workers AICloud$0.35$0.75128K
PrivatemodeHost$0.231$0.751128K
STACKITCloud$0.53$0.76131K
RegoloHost$1.12$4.71128K

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Run locally

116.83B parameters
16-bit
246.3 GB
8-bit
131.3 GB
4-bit
75.4 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 128 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.6 counted40.1178 / 218
CodingCode writing and repair.4 counted42.8173 / 209
ReasoningLogic problems and puzzles.3 counted40.2164 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.5 counted42.4181 / 268
MultilingualTasks in many languages.1 counted51.1—
InstructionTasks with strict prompt rules.1 counted53.5—
MathMath problems.5 counted51.579 / 168

Results

25 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
American Invitational Mathematics Examination 2024Math95.8%2 / 27
MATH 500Math94.8% ±1.06 / 59
EuroEval PolishMultilingual57.6% ±0.527 / 209
EuroEval SwedishMultilingual61.3% ±0.930 / 195
EuroEval DutchMultilingual57.8% ±0.430 / 175
EuroEval GermanMultilingual51.2% ±0.633 / 191
EuroEval PortugueseMultilingual59.3% ±0.532 / 182
EuroEval SpanishMultilingual53.5% ±0.431 / 176
Massive Multitask Language UnderstandingKnowledge90.0%6 / 34
EuroEval ItalianMultilingual58.4% ±0.632 / 174
AIMEMath92.6% ±1.218 / 95
American Invitational Mathematics Examination 2025Math92.5%13 / 53
Artificial Analysis LiveCodeBenchCodingAbove fitFar above this model's index.87.8%3 / 12
EuroEval FrenchMultilingual58.4% ±0.445 / 178
OTIS Mock AIME 2024-2025MathAbove fitFar above this model's index.88.9% ±4.563 / 200
Artificial Analysis IFBenchInstruction69.0%48 / 146
LiveCodeBenchCoding83.2% ±1.149 / 142
Graduate-Level Google-Proof Q&AKnowledge80.1%66 / 141
GPQA diamond · Epoch AIKnowledge75.8% ±2.7103 / 217
Artificial Analysis Humanity's Last ExamKnowledge19.6%113 / 222
Artificial Analysis GPQA DiamondKnowledge78.2%105 / 199
GPQA Diamond · Vals AIKnowledge78.5% ±2.372 / 135
τ²-Bench Tool-Agent-User EvaluationAgentic65.8%81 / 149
Critical Physics TasksReasoning1.1%118 / 214
Artificial Analysis Long Context ReasoningReasoning52.0%152 / 217
GDPval-AA normalizedAgentic5.4%96 / 133
MMLU ProKnowledge79.2% ±0.498 / 135
Software Engineering Benchmark VerifiedCoding62.4%83 / 114
SWE-bench Verified · swebench.comCoding26.0%34 / 46
Artificial Analysis Omniscience AccuracyKnowledge21.7%122 / 164
Terminal-Bench 1.0Agentic22.5% ±4.737 / 47
React Native EvalsCoding71.6%13 / 16
SWE-bench Verified · SWE-bench/experimentsCoding26.0%56 / 65
Terminal-Bench 2.0 · Vals AIAgentic19.1% ±4.258 / 66
Artificial Analysis SciCodeCoding34.0%101 / 114
Humanity's Last ExamKnowledge14.9%71 / 77
APEX-Agents-AAAgentic3.1%28 / 30
Gert Labs Composite Game BenchmarkAgentic29.6%49 / 52
SWE-benchCoding33.6% ±2.186 / 87
Mystery Game PuzzlesReasoning0.0%80 / 80
MGSMMultilingual92.0% ±1.624 / 74
MedQAHealthcare91.4% ±0.238 / 95
TaxEval v2Finance71.6% ±0.976 / 143
CorpFin v2Finance58.2% ±1.084 / 131
LegalBenchLegal75.9% ±0.5113 / 149
Finance Agent v1.1Finance21.5% ±2.245 / 51
CaseLaw v2Legal48.8% ±0.949 / 53
Artificial Analysis AIME 2025Math93.4%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseVals AINo licence statedEpoch AICC BYEuroEvalMITSWE-bench teamNo licence statedmodels.devMIT

History

5 releases
Index44.0 → 43.4
23 Oct 202612 Nov 2026
Output / 1M$0.60 → $0.60
23 Oct 202612 Nov 2026

Questions

How much does GPT-OSS 120B cost?
$0.15 per 1M input tokens, $0.60 per 1M output tokens.
How good is GPT-OSS 120B?
Its index is 43.4 out of 100. That is place 189 of 283 ranked models.
What is GPT-OSS 120B best at?
Math: 51.5 out of 100, place 79 of 168.
How fast is GPT-OSS 120B?
34 output tokens per second, 0.3 s to the first token.
What is the context window of GPT-OSS 120B?
131K tokens.
Is GPT-OSS 120B open source?
Yes. Its weights are published.
When was GPT-OSS 120B released?
August 2025.

Badge

GPT-OSS 120B on Inference360: place 189 of 283, index 43.4
HTML
<a href="https://inference360.ai/models/gpt-oss-120b"><img src="https://inference360.ai/badges/gpt-oss-120b.svg" alt="GPT-OSS 120B on Inference360: place 189 of 283, index 43.4" height="20"></a>
Markdown
[![GPT-OSS 120B on Inference360: place 189 of 283, index 43.4](https://inference360.ai/badges/gpt-oss-120b.svg)](https://inference360.ai/models/gpt-oss-120b)

Compare with

MiMo-V2-Flash43.4Trinity-Large-Thinking43.5Kimi K2 Instruct43.2Claude Haiku 4.543.2o4-mini44.7

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeNemotron 3 Ultra50.8 · Free

More from OpenAI

All models
GPT-OSS 20B41.7GPT-5 (high)53.6GPT-5 (medium)53.0GPT-5 mini48.4GPT-5 nano41.9