Compare · ModelsLive · 2 picked · head to head
Gemini 2.5 Pro vs Gemma 4 26B A4B
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Gemini 2.5 Pro wins on 8/9 benchmarks
Gemini 2.5 Pro wins 8 of 9 shared benchmarks. Leads in speed · knowledge · general.
Category leads
speed·Gemini 2.5 Proarena·Gemma 4 26B A4B knowledge·Gemini 2.5 Progeneral·Gemini 2.5 Promath·Gemini 2.5 Procoding·Gemini 2.5 Pro
Hype vs Reality
Attention vs performance
Gemini 2.5 Pro
#116 by perf·no signal
Gemma 4 26B A4B
#151 by perf·#7 by attention
Best value
Gemma 4 26B A4B
34.8x better value than Gemini 2.5 Pro
Gemini 2.5 Pro
9.0 pts/$
$5.63/M
Gemma 4 26B A4B
314.5 pts/$
$0.15/M
Vendor risk
Who is behind the model
Google DeepMind
$4.20T·Tier 1
Google DeepMind
$4.20T·Tier 1
Head to head
9 benchmarks · 2 models
Gemini 2.5 ProGemma 4 26B A4B
Artificial Analysis · Quality Index
Gemini 2.5 Pro leads by +10.3
Gemini 2.5 Pro
27.0
Gemma 4 26B A4B
16.7
Chatbot Arena Elo · Coding
Gemma 4 26B A4B leads by +131.5
Gemini 2.5 Pro
1227.0
Gemma 4 26B A4B
1358.5
Chatbot Arena Elo · Overall
Gemini 2.5 Pro leads by +8.1
Gemini 2.5 Pro
1445.6
Gemma 4 26B A4B
1437.5
Chess Puzzles
Gemini 2.5 Pro leads by +14.7
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
Gemma 4 26B A4B
1.1
Dtbench
Gemini 2.5 Pro leads by +12.5
Gemini 2.5 Pro
70.7
Gemma 4 26B A4B
58.2
GPQA diamond
Gemini 2.5 Pro leads by +16.1
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
Gemma 4 26B A4B
64.3
Lmca
Gemini 2.5 Pro leads by +6.0
Gemini 2.5 Pro
40.9
Gemma 4 26B A4B
34.9
OTIS Mock AIME 2024-2025
Gemini 2.5 Pro leads by +2.5
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Gemini 2.5 Pro
84.7
Gemma 4 26B A4B
82.2
WeirdML
Gemini 2.5 Pro leads by +18.9
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Gemini 2.5 Pro
54.0
Gemma 4 26B A4B
35.2
Full benchmark table
| Benchmark | Gemini 2.5 Pro | Gemma 4 26B A4B |
|---|---|---|
Artificial Analysis · Quality Index | 27.0 | 16.7 |
Chatbot Arena Elo · Coding | 1227.0 | 1358.5 |
Chatbot Arena Elo · Overall | 1445.6 | 1437.5 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 15.8 | 1.1 |
Dtbench | 70.7 | 58.2 |
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 | 64.3 |
Lmca | 40.9 | 34.9 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 84.7 | 82.2 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 54.0 | 35.2 |
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 | |
| $0.07 | $0.23 | 262K tokens (~131 books) | $1.07 |