Compare · ModelsLive · 2 picked · head to head
Gemini 2.5 Pro vs gpt-oss-20b
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Gemini 2.5 Pro wins on 19/20 benchmarks
Gemini 2.5 Pro wins 19 of 20 shared benchmarks. Leads in speed · arena · knowledge.
Category leads
speed·Gemini 2.5 Proarena·Gemini 2.5 Proknowledge·Gemini 2.5 Progeneral·Gemini 2.5 Prolanguage·Gemini 2.5 Promath·Gemini 2.5 Proreasoning·Gemini 2.5 Procoding·Gemini 2.5 Pro
Hype vs Reality
Attention vs performance
Gemini 2.5 Pro
#116 by perf·no signal
gpt-oss-20b
#97 by perf·no signal
Best value
gpt-oss-20b
108.9x better value than Gemini 2.5 Pro
Gemini 2.5 Pro
9.0 pts/$
$5.63/M
gpt-oss-20b
983.3 pts/$
$0.05/M
Vendor risk
Who is behind the model
Google DeepMind
$4.20T·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
20 benchmarks · 2 models
Gemini 2.5 Progpt-oss-20b
Artificial Analysis · Quality Index
Gemini 2.5 Pro leads by +18.0
Gemini 2.5 Pro
27.0
gpt-oss-20b
9.0
Chatbot Arena Elo · Overall
Gemini 2.5 Pro leads by +128.0
Gemini 2.5 Pro
1445.6
gpt-oss-20b
1317.6
Chess Puzzles
Gemini 2.5 Pro leads by +15.8
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Gemini 2.5 Pro
15.8
gpt-oss-20b
0.0
Dtbench
Gemini 2.5 Pro leads by +24.0
Gemini 2.5 Pro
70.7
gpt-oss-20b
46.7
GPQA diamond
Gemini 2.5 Pro leads by +32.7
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Gemini 2.5 Pro
80.4
gpt-oss-20b
47.7
HELM · GPQA
Gemini 2.5 Pro leads by +15.5
Gemini 2.5 Pro
74.9
gpt-oss-20b
59.4
HELM · IFEval
Gemini 2.5 Pro leads by +10.8
Gemini 2.5 Pro
84.0
gpt-oss-20b
73.2
HELM · MMLU-Pro
Gemini 2.5 Pro leads by +12.3
Gemini 2.5 Pro
86.3
gpt-oss-20b
74.0
HELM · Omni-MATH
gpt-oss-20b leads by +14.9
Gemini 2.5 Pro
41.6
gpt-oss-20b
56.5
HELM · WildBench
Gemini 2.5 Pro leads by +12.0
Gemini 2.5 Pro
85.7
gpt-oss-20b
73.7
Lmca
Gemini 2.5 Pro leads by +23.8
Gemini 2.5 Pro
40.9
gpt-oss-20b
17.1
OpenCompass · AIME2025
Gemini 2.5 Pro leads by +0.8
Gemini 2.5 Pro
88.7
gpt-oss-20b
87.9
OpenCompass · GPQA-Diamond
Gemini 2.5 Pro leads by +15.8
Gemini 2.5 Pro
84.7
gpt-oss-20b
68.9
OpenCompass · HLE
Gemini 2.5 Pro leads by +9.5
Gemini 2.5 Pro
21.1
gpt-oss-20b
11.6
OpenCompass · IFEval
Gemini 2.5 Pro leads by +1.1
Gemini 2.5 Pro
90.0
gpt-oss-20b
88.9
OpenCompass · LiveCodeBenchV6
Gemini 2.5 Pro leads by +2.9
Gemini 2.5 Pro
71.3
gpt-oss-20b
68.4
OpenCompass · MMLU-Pro
Gemini 2.5 Pro leads by +13.0
Gemini 2.5 Pro
85.8
gpt-oss-20b
72.8
OTIS Mock AIME 2024-2025
Gemini 2.5 Pro leads by +19.5
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Gemini 2.5 Pro
84.7
gpt-oss-20b
65.2
Terminal Bench
Gemini 2.5 Pro leads by +29.2
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.
Gemini 2.5 Pro
32.6
gpt-oss-20b
3.4
WeirdML
Gemini 2.5 Pro leads by +13.1
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Gemini 2.5 Pro
54.0
gpt-oss-20b
40.9
Full benchmark table
| Benchmark | Gemini 2.5 Pro | gpt-oss-20b |
|---|---|---|
Artificial Analysis · Quality Index | 27.0 | 9.0 |
Chatbot Arena Elo · Overall | 1445.6 | 1317.6 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 15.8 | 0.0 |
Dtbench | 70.7 | 46.7 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 80.4 | 47.7 |
HELM · GPQA | 74.9 | 59.4 |
HELM · IFEval | 84.0 | 73.2 |
HELM · MMLU-Pro | 86.3 | 74.0 |
HELM · Omni-MATH | 41.6 | 56.5 |
HELM · WildBench | 85.7 | 73.7 |
Lmca | 40.9 | 17.1 |
OpenCompass · AIME2025 | 88.7 | 87.9 |
OpenCompass · GPQA-Diamond | 84.7 | 68.9 |
OpenCompass · HLE | 21.1 | 11.6 |
OpenCompass · IFEval | 90.0 | 88.9 |
OpenCompass · LiveCodeBenchV6 | 71.3 | 68.4 |
OpenCompass · MMLU-Pro | 85.8 | 72.8 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 84.7 | 65.2 |
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. | 32.6 | 3.4 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 54.0 | 40.9 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $1.25 | $10.00 | 1.0M tokens (~524 books) | $34.38 | |
| $0.02 | $0.09 | 131K tokens (~66 books) | $0.36 |