vs
14:1
overall capability wins (2 ties across 17 scored). Gemma 3 27B leads overall.
- Reasoning8:1
- Coding3:0
- Arena1:0
- Tool use1:0
- Pricing0:3
- Speed0:2
- Specs1:0
Verdict
Gemma 3 27B leads coding (widest gap: +17.1 on LiveCode); Gemma 3 27B ranks higher on LMArena (+62 Elo); Gemma 3 4B is ~3.0x cheaper on a blended token basis than Gemma 3 27B.
- Gemma 3 27B leads coding (widest gap: +17.1 on LiveCode)
- Gemma 3 27B ranks higher on LMArena (+62 Elo)
- Gemma 3 4B is ~3.0x cheaper on a blended token basis than Gemma 3 27B
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Gemma 3 27B
- $0.17
- Gemma 3 4B
- $0.06
Gemma 3 4B is about 2.9x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Family
Gemma
Gemma
License
Gemma
Gemma
Open weights
Yes
Yes
Release
Mar 12, 2025
Mar 12, 2025
Knowledge cutoff
-
-
API / provider
Google AI / self-host
Together / Fireworks (ref.)
Modalities
text, image → text
text, image → text
Specs
Context window
128K
128K
Max output
8K
8K
Parameters
27B
4B
Pricing
Input $/1M
$0.10
$0.04
Output $/1M
$0.30
$0.08
Blended $/1M (3∶1)
$0.15
$0.05
Speed
tok/s
90
160
TTFT (s)
0.28
0.18
Reasoning
MMLU-Pro
67.5
43.6
GPQA Diamond
42.8
29.1
Humanity's Last Exam
4.4
5.3
AIME 2025
20.7
12.7
MATH-500
88.3
76.6
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
13.0
7.7
AA-LCR v1.1
7.3
6.7
CritPt
—
—
MMMU-Pro
48.0
29.9
IFBench
31.8
28.3
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
11.4
—
SWE-bench Multilingual
—
—
LiveCodeBench
29.7
12.6
Terminal-Bench 2.1
4.5
0.4
Aider Polyglot
4.9
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
23.3
7.3
CursorBench
—
—
SWE-Rebench
5.4
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,365
1,303
| Metric | Gemma 3 27B | Gemma 3 4B | Delta |
|---|---|---|---|
| Identity | |||
| Organization | — | ||
| Family | Gemma | Gemma | — |
| License | Gemma | Gemma | — |
| Open weights | Yes | Yes | — |
| Release | Mar 12, 2025 | Mar 12, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Google AI / self-host | Together / Fireworks (ref.) | — |
| Modalities | text, image → text | text, image → text | — |
| Specs | |||
| Context window | 128K | 128K | tie |
| Max output | 8K | 8K | tie |
| Parameters | 27B | 4B | A +23 |
| Pricing | |||
| Input $/1M | $0.10 | $0.04 | B +$0.06 |
| Output $/1M | $0.30 | $0.08 | B +$0.22 |
| Blended $/1M (3∶1) | $0.15 | $0.05 | B +$0.10 |
| Speed | |||
| tok/s | 90 | 160 | B +70 |
| TTFT (s) | 0.28 | 0.18 | B +0.10 |
| Reasoning | |||
| MMLU-Pro | 67.5 | 43.6 | A +23.9 pts |
| GPQA Diamond | 42.8 | 29.1 | A +13.7 pts |
| Humanity's Last Exam | 4.4 | 5.3 | B +0.9 pts |
| AIME 2025 | 20.7 | 12.7 | A +8.0 pts |
| MATH-500 | 88.3 | 76.6 | A +11.7 pts |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 13.0 | 7.7 | A +5.3 pts |
| AA-LCR v1.1 | 7.3 | 6.7 | A +0.6 pts |
| CritPt | — | — | — |
| MMMU-Pro | 48.0 | 29.9 | A +18.1 pts |
| IFBench | 31.8 | 28.3 | A +3.5 pts |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | 11.4 | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 29.7 | 12.6 | A +17.1 pts |
| Terminal-Bench 2.1 | 4.5 | 0.4 | A +4.1 pts |
| Aider Polyglot | 4.9 | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 23.3 | 7.3 | A +16.0 pts |
| CursorBench | — | — | — |
| SWE-Rebench | 5.4 | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,365 | 1,303 | A +62 Elo |
Gemma 3 27B
- Context
- 128K / 8K out
- Parameters
- 27B
- Price
- $0.10 / $0.30
- Speed
- 90 tok/s · 0.28s TTFT
- Modalities
- text, image → text
- License
- Gemma
Open multimodal Gemma 3 mid-size - strong preference Elo for its parameter count.
Gemma 3 4B
- Context
- 128K / 8K out
- Parameters
- 4B
- Price
- $0.04 / $0.08
- Speed
- 160 tok/s · 0.18s TTFT
- Modalities
- text, image → text
- License
- Gemma
Small multimodal Gemma 3 with long context — strong on-device and edge deployment option.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.