Compare · ModelsLive · 2 picked · head to head
gpt-oss-120b vs GPT-5 Nano
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
gpt-oss-120b wins on 12/20 benchmarks
gpt-oss-120b wins 12 of 20 shared benchmarks. Leads in knowledge · math · reasoning.
Category leads
knowledge·gpt-oss-120blanguage·GPT-5 Nanomath·gpt-oss-120breasoning·gpt-oss-120bcoding·GPT-5 Nano
Hype vs Reality
Attention vs performance
gpt-oss-120b
#137 by perf·no signal
GPT-5 Nano
#153 by perf·no signal
Best value
gpt-oss-120b
2.3x better value than GPT-5 Nano
gpt-oss-120b
434.3 pts/$
$0.11/M
GPT-5 Nano
192.4 pts/$
$0.23/M
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
20 benchmarks · 2 models
gpt-oss-120bGPT-5 Nano
Fiction.LiveBench
Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination.
gpt-oss-120b
44.4
GPT-5 Nano
44.4
GPQA diamond
gpt-oss-120b leads by +8.4
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
gpt-oss-120b
67.7
GPT-5 Nano
59.3
HELM · GPQA
gpt-oss-120b leads by +0.5
gpt-oss-120b
68.4
GPT-5 Nano
67.9
HELM · IFEval
GPT-5 Nano leads by +9.6
gpt-oss-120b
83.6
GPT-5 Nano
93.2
HELM · MMLU-Pro
gpt-oss-120b leads by +1.7
gpt-oss-120b
79.5
GPT-5 Nano
77.8
HELM · Omni-MATH
gpt-oss-120b leads by +14.1
gpt-oss-120b
68.8
GPT-5 Nano
54.7
HELM · WildBench
gpt-oss-120b leads by +3.9
gpt-oss-120b
84.5
GPT-5 Nano
80.6
LiveBench · Agentic Coding
GPT-5 Nano leads by +11.7
gpt-oss-120b
16.7
GPT-5 Nano
28.3
LiveBench · Coding
GPT-5 Nano leads by +7.2
gpt-oss-120b
60.2
GPT-5 Nano
67.4
LiveBench · Data Analysis
GPT-5 Nano leads by +5.5
gpt-oss-120b
38.8
GPT-5 Nano
44.3
LiveBench · If
GPT-5 Nano leads by +1.7
gpt-oss-120b
50.3
GPT-5 Nano
52.0
LiveBench · Language
gpt-oss-120b leads by +0.9
gpt-oss-120b
48.6
GPT-5 Nano
47.7
LiveBench · Mathematics
gpt-oss-120b leads by +4.2
gpt-oss-120b
68.9
GPT-5 Nano
64.7
LiveBench · Overall
GPT-5 Nano leads by +2.5
gpt-oss-120b
46.1
GPT-5 Nano
48.6
LiveBench · Reasoning
gpt-oss-120b leads by +3.8
gpt-oss-120b
39.2
GPT-5 Nano
35.5
OTIS Mock AIME 2024-2025
gpt-oss-120b leads by +7.8
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
gpt-oss-120b
88.9
GPT-5 Nano
81.1
SimpleQA Verified
gpt-oss-120b leads by +1.7
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
gpt-oss-120b
13.9
GPT-5 Nano
12.2
SWE-Bench Verified (Bash Only)
GPT-5 Nano leads by +8.8
SWE-Bench Verified (Bash Only) · a curated subset of SWE-bench where models fix real Python repository bugs using only bash commands, no agent frameworks.
gpt-oss-120b
26.0
GPT-5 Nano
34.8
Terminal Bench
GPT-5 Nano leads by +3.1
Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence.
gpt-oss-120b
18.7
GPT-5 Nano
21.8
WeirdML
gpt-oss-120b leads by +10.1
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
gpt-oss-120b
48.2
GPT-5 Nano
38.1
Full benchmark table
| Benchmark | gpt-oss-120b | GPT-5 Nano |
|---|---|---|
Fiction.LiveBench Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination. | 44.4 | 44.4 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 67.7 | 59.3 |
HELM · GPQA | 68.4 | 67.9 |
HELM · IFEval | 83.6 | 93.2 |
HELM · MMLU-Pro | 79.5 | 77.8 |
HELM · Omni-MATH | 68.8 | 54.7 |
HELM · WildBench | 84.5 | 80.6 |
LiveBench · Agentic Coding | 16.7 | 28.3 |
LiveBench · Coding | 60.2 | 67.4 |
LiveBench · Data Analysis | 38.8 | 44.3 |
LiveBench · If | 50.3 | 52.0 |
LiveBench · Language | 48.6 | 47.7 |
LiveBench · Mathematics | 68.9 | 64.7 |
LiveBench · Overall | 46.1 | 48.6 |
LiveBench · Reasoning | 39.2 | 35.5 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 88.9 | 81.1 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 13.9 | 12.2 |
SWE-Bench Verified (Bash Only) SWE-Bench Verified (Bash Only) · a curated subset of SWE-bench where models fix real Python repository bugs using only bash commands, no agent frameworks. | 26.0 | 34.8 |
Terminal Bench Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence. | 18.7 | 21.8 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 48.2 | 38.1 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.04 | $0.18 | 131K tokens (~66 books) | $0.72 | |
| $0.05 | $0.40 | 400K tokens (~200 books) | $1.38 |
People also compared