#28 Gemma 4 26B A4B
medium
Invalid SVG
- Cost
- $0.000
- Time
- 300.0s
- Tokens
- 0 tok
Google
Release: 2026-04-03
Tested on: 2026-04-22 21:39
google/gemma-4-26b-a4b-it::medium
25.2B total (3.8B active)MoEOpen source
(medium)
(none)
Summary
Gemma 4 26B A4B scores 8.0 on AI BENCHY and ranks #28. It has N/A reliability, a 75.9% pass rate, $0.028 total cost, and 25.03s average response time.
What makes Gemma 4 26B A4B unique: Its total benchmark cost is unusually low for its score range.
Model facts
Researched on 2026-08-12
- Parameters
- 25.2B total (3.8B active)
- Architecture
- MoE
- Availability
- Open source
- License
- Apache-2.0
8.0
Consistency
9.0
N/A
$0.028
Total Output Tokens
60,559
Total Input Tokens
0
Input Price
$0.070 / 1M
Output Price
$0.340 / 1M
Flaky tests
2
Flaky tests had mixed outcomes across runs (at least one pass and one fail).
Hamster playing table tennis
Prompt: Create a detailed SVG illustration of a hamster playing table tennis.
Run history
| Tested on | Score | Reliability | Tests Correct | Total Cost | Compare |
|---|---|---|---|---|---|
| 2026-08-14 02:53 Re-test | 6.8 | 10.0 | $0.111 ↑ | Compare | |
| 2026-07-16 23:43 New test added | 6.6 | 9.4 | $0.082 ↓ | Compare | |
| 2026-06-04 14:22 New test added | 7.6 | 8.9 | $0.045 ↓ | Compare | |
| 2026-05-22 00:42 Suite changed | 7.5 | 10.0 | $0.045 | Compare | |
| 2026-04-22 21:39 First recorded run | 8.0 | N/A | $0.028 | Current run |
Run comparison
| Run | Benchmark coverage | Score | Consistency | Reliability | Tests Correct | Flaky tests | Total Output Tokens | Total Input Tokens | Total Cost | Response Time (avg) |
|---|---|---|---|---|---|---|---|---|---|---|
| 2026-04-22 21:39 · First recorded run | 54/54 attempts | 8.0 | 9.0 | N/A | 13/18 | 2 | 60,559 | 0 | $0.028 | 25.03s |
| 2026-06-04 14:22 · New test added | 63/63 attempts | 7.6 | 9.2 | 8.9 | 14/21 | 2 | 128,490 | 40,252 | $0.045 | 63.41s |
| Difference | — | +0.5 | -0.2 | -1 | 0 | -67931 | -40252 | -$0.017 | -38386ms |
Benchmark coverage differs: 54/54 attempts (Target: 3 repeats per test) versus 63/63 attempts (Target: 3 repeats per test). Totals and repeat-sensitive metrics are not directly comparable.
These two runs used different benchmark suites, so the deltas reflect both model changes and suite changes.
Charts
Choose the first model, then click a second model to open a side-by-side page.
Top Models by Score
Score vs Total Cost
Response Time (avg)
Score vs Response Time (avg)
Total Output Tokens
Score vs Total Output Tokens
Quick Compare
Gemma 4 26B A4BmediumFree AvailablevsGemini 3 Flash PreviewnoneGemma 4 26B A4BmediumFree AvailablevsGLM 5.2noneGemma 4 26B A4BmediumFree AvailablevsMiMo-V2.5-PromediumGemma 4 26B A4BmediumFree AvailablevsKAT-Coder-Pro V2.5noneGemma 4 26B A4BmediumFree AvailablevsDeepSeek V4 PrononeGemma 4 26B A4BmediumFree AvailablevsLongCat 2.0lowGemma 4 26B A4BmediumFree AvailablevsSolar Pro 4mediumGemma 4 26B A4BmediumFree AvailablevsGemini 3.5 FlashminimalGemma 4 26B A4BmediumFree AvailablevsSolar Pro 4lowGemma 4 26B A4BmediumFree AvailablevsLongCat 2.0high
Category Breakdown
| Category | Score | Consistency | Tests Correct |
|---|---|---|---|
| Anti-AI Tricks | 10.0 | 10.0 | |
| Coding | 2.8 | 10.0 | |
| Combined | 9.6 | 10.0 | |
| Data parsing and extraction | 10.0 | 10.0 | |
| Domain specific | 2.9 | 4.4 | |
| General Intelligence | 10.0 | 10.0 | |
| Instructions following | 10.0 | 10.0 | |
| Puzzle Solving | 7.9 | 9.6 | |
| Tool Calling | 10.0 | 10.0 |