OpenAI · Benchmarks

gpt-oss-120b benchmarks

gpt-oss-120b has 5 published scores in this dataset, from a OpenAI release dated 2025-08-05. Its composite index is 57.3, which places it 22nd of 22 ranked models and 15th among 15 open-weight models. Its strongest area is coding (index 76.5) and its weakest is agentic (index 27.1). Its best placing is on AIME, where its 95.8 ranks 2nd of 5 models with a published score.

Open weightsOfficial page

Category profile

Index 0–100: each score relative to the best published score on that test.

ReasoningCodingAgenticKnowledge
Reasoning
68.2#14 of 16 · 3 tests
Coding
76.51 test · not ranked
Agentic
27.11 test · not ranked
Knowledge
—no data

Model facts

Lab
OpenAI
Type
Open weights
Released
2025-08-05
Parameters
117B total, 5.1B active (MoE)
Context
128K
License
Apache 2.0

gpt-oss-120b hardware requirements and cost

Even at 4-bit it needs about 81 GB (1× H200 (141 GB)), well beyond a home workstation, so most people run it on rented multi-GPU servers. Full BF16 precision needs about 269 GB (4× H100 (320 GB)).

Estimated GPU memory for gpt-oss-120b
PrecisionWeightsWith headroomSmallest setup
BF16 (full precision)234 GB269 GB4× H100 (320 GB)
FP8 / 8-bit117 GB135 GB1× H200 (141 GB)
4-bit (GGUF Q4_K_M)70 GB81 GB1× H200 (141 GB)

Estimate: parameters × bytes per parameter, plus 15% for the KV cache and runtime at modest context lengths. Long contexts and large batches need more.

Lighter or stronger open models

Open models ranked just above and below on the composite index, with their 4-bit memory needs.

gpt-oss-120b speed and pricing

Measured figures link to their source; estimates are labelled.

API output speed
212 tok/sArtificial Analysis median · Artificial Analysis
Time to first token
0.53 sOpenRouter model page provider stats
API price per 1M tokens
$0.030 in · $0.150 outcheapest provider on OpenRouter: Near AI · OpenRouter endpoints API
Compare speed and price across all models

Score by test

Rank is among every model with a published score on that test.

GPQA Diamond80.1%high effort, no tools#19 of 19 · 16.0 behind GPT-6 AstraHumanity's Last Exam14.9%high effort, no tools#22 of 23 · 49.5 behind Claude Opus 5.5Humanity's Last Exam (with tools)19.0%high effort, with tools#16 of 17 · 51.0 behind Kimi K3.1 ProAIME95.8%AIME 2024, high effort, no tools#2 of 5 · 1.9 behind Reflection Beam 1.0SWE-bench Verified62.4%high effort#5 of 5 · 19.2 behind Reflection Beam 1.0

Nearby in the composite ranking

Models with the closest composite index.

  1. 1Gemma 4 26B A4B72.9
  2. 2Claude Haiku 5.572.4
  3. 3Gemini 3.8 Flash70.8
  4. 4gpt-oss-120b57.3

Head-to-head comparisons

Side-by-side pages built from the tests both models report.

Sources

ModelTestScoreSettingSource
gpt-oss-120bGPQA Diamond80.1high effort, no toolsgpt-oss model card paper (arXiv 2508.10925), Table 3
gpt-oss-120bHumanity's Last Exam14.9high effort, no toolsgpt-oss model card paper (arXiv 2508.10925), Table 3
gpt-oss-120bHumanity's Last Exam (with tools)19.0high effort, with toolsgpt-oss model card paper (arXiv 2508.10925), Table 3
gpt-oss-120bAIME95.8AIME 2024, high effort, no toolsgpt-oss model card paper (arXiv 2508.10925), Table 3
gpt-oss-120bSWE-bench Verified62.4high effortgpt-oss model card paper (arXiv 2508.10925), Table 3

Scores are published results from model cards, lab reports, and public leaderboards, each linked to its source. Settings such as thinking mode or test version differ between sources and are shown next to each score. OpenLLM Buddy catalog models are uncensored fine-tunes; their scores are those of the official base model. Data collected 2026-10-30. How the index works