Compare · ModelsLive · 2 picked · head to head
gpt-oss-120b vs o3
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
o3 wins on 26/28 benchmarks
o3 wins 26 of 28 shared benchmarks. Leads in speed · coding · agentic.
Category leads
speed·o3coding·o3agentic·o3knowledge·o3general·o3language·o3math·o3reasoning·o3
Hype vs Reality
Attention vs performance
gpt-oss-120b
#127 by perf·no signal
o3
#111 by perf·no signal
Best value
gpt-oss-120b
46.5x better value than o3
gpt-oss-120b
475.4 pts/$
$0.10/M
o3
10.2 pts/$
$5.00/M
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
28 benchmarks · 2 models
gpt-oss-120bo3
Artificial Analysis · CritPt
gpt-oss-120b
1.1
o3
1.1
Artificial Analysis · GPQA Diamond
o3 leads by +4.5
gpt-oss-120b
78.2
o3
82.7
Artificial Analysis · Humanity's Last Exam
o3 leads by +0.5
gpt-oss-120b
19.6
o3
20.1
Artificial Analysis · IFBench
o3 leads by +2.4
gpt-oss-120b
69.0
o3
71.4
Artificial Analysis · Long Context Reasoning
o3 leads by +22.7
gpt-oss-120b
52.0
o3
74.7
Artificial Analysis · Quality Index
o3 leads by +8.6
gpt-oss-120b
11.6
o3
20.2
Artificial Analysis · tau2-Bench Telecom
o3 leads by +14.9
gpt-oss-120b
65.8
o3
80.7
Artificial Analysis · Terminal-Bench Hard
o3 leads by +13.6
gpt-oss-120b
23.5
o3
37.1
Aider polyglot
o3 leads by +39.5
Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework.
gpt-oss-120b
41.8
o3
81.3
APEX-Agents
o3 leads by +12.8
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
gpt-oss-120b
4.4
o3
17.2
Chess Puzzles
o3 leads by +18.9
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
gpt-oss-120b
15.8
o3
34.8
Dtbench
o3 leads by +14.2
gpt-oss-120b
60.5
o3
74.7
Fiction.LiveBench
o3 leads by +44.5
Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination.
gpt-oss-120b
44.4
o3
88.9
GPQA diamond
o3 leads by +8.1
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
o3
75.8
HELM · GPQA
o3 leads by +6.9
gpt-oss-120b
68.4
o3
75.3
HELM · IFEval
o3 leads by +3.3
gpt-oss-120b
83.6
o3
86.9
HELM · MMLU-Pro
o3 leads by +6.4
gpt-oss-120b
79.5
o3
85.9
HELM · Omni-MATH
o3 leads by +2.6
gpt-oss-120b
68.8
o3
71.4
HELM · WildBench
o3 leads by +1.6
gpt-oss-120b
84.5
o3
86.1
Lech Mazur Writing
o3 leads by +6.8
Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication.
gpt-oss-120b
77.1
o3
83.9
Lmca
o3 leads by +20.7
gpt-oss-120b
26.1
o3
46.7
Metr Time Horizons
o3 leads by +8.8
gpt-oss-120b
56.6
o3
65.4
Mystery Game Puzzles
o3 leads by +21.8
gpt-oss-120b
0.0
o3
21.8
OTIS Mock AIME 2024-2025
gpt-oss-120b leads by +4.4
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
gpt-oss-120b
88.9
o3
84.4
SimpleBench
o3 leads by +37.2
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
gpt-oss-120b
6.5
o3
43.7
SimpleQA Verified
o3 leads by +35.5
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
o3
49.4
SWE-Bench Verified (Bash Only)
o3 leads by +32.4
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
o3
58.4
WeirdML
o3 leads by +4.3
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
gpt-oss-120b
48.2
o3
52.4
Full benchmark table
| Benchmark | gpt-oss-120b | o3 |
|---|---|---|
Artificial Analysis · CritPt | 1.1 | 1.1 |
Artificial Analysis · GPQA Diamond | 78.2 | 82.7 |
Artificial Analysis · Humanity's Last Exam | 19.6 | 20.1 |
Artificial Analysis · IFBench | 69.0 | 71.4 |
Artificial Analysis · Long Context Reasoning | 52.0 | 74.7 |
Artificial Analysis · Quality Index | 11.6 | 20.2 |
Artificial Analysis · tau2-Bench Telecom | 65.8 | 80.7 |
Artificial Analysis · Terminal-Bench Hard | 23.5 | 37.1 |
Aider polyglot Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework. | 41.8 | 81.3 |
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 4.4 | 17.2 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 15.8 | 34.8 |
Dtbench | 60.5 | 74.7 |
Fiction.LiveBench Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination. | 44.4 | 88.9 |
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 | 75.8 |
HELM · GPQA | 68.4 | 75.3 |
HELM · IFEval | 83.6 | 86.9 |
HELM · MMLU-Pro | 79.5 | 85.9 |
HELM · Omni-MATH | 68.8 | 71.4 |
HELM · WildBench | 84.5 | 86.1 |
Lech Mazur Writing Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication. | 77.1 | 83.9 |
Lmca | 26.1 | 46.7 |
Metr Time Horizons | 56.6 | 65.4 |
Mystery Game Puzzles | 0.0 | 21.8 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 88.9 | 84.4 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 6.5 | 43.7 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 13.9 | 49.4 |
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 | 58.4 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 48.2 | 52.4 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.04 | $0.17 | 131K tokens (~66 books) | $0.70 | |
| $2.00 | $8.00 | 200K tokens (~100 books) | $35.00 |