Compare/Granite 4.0 H Small vs Gemma 3 27B Instruct

Granite 4.0 H SmallvsGemma 3 27B Instruct

Side-by-side comparison of pricing, 12 benchmarks, and generation speed.

IBM

Granite 4.0 H Small

Input
$0.06/M
Output
$0.25/M
Speed
265 tok/s
TTFT
8.78s
Google

Gemma 3 27B Instruct

Input
$0.11/M
Output
$0.25/M
Speed
TTFT

Winner by Category

Cheaper
Granite 4.0 H Small
Faster (tok/s)
Granite 4.0 H Small
Lower Latency
Gemma 3 27B Instruct
Benchmarks (2-10)
Gemma 3 27B Instruct

Pricing Comparison

MetricGranite 4.0 H SmallGemma 3 27B Instruct
Input ($/M tokens)$0.06$0.11
Output ($/M tokens)$0.25$0.25
Cost for 1M input + 100K output tokens:
Granite 4.0 H Small$0.08
Gemma 3 27B Instruct$0.14

Speed Comparison

Output Speed (tokens/s) — higher is better
Granite 4.0 H Small
265 tok/s
Gemma 3 27B Instruct
Time to First Token (seconds) — lower is better
Granite 4.0 H Small
8.78s
Gemma 3 27B Instruct

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
10.810.3
Coding Index
8.59.6
Math Index
13.720.7
GPQA Diamond
41.6%42.8%
MMLU-Pro
62.4%66.9%
LiveCodeBench
25.1%13.7%
AIME 2025
13.7%20.7%
MATH-500
88.3%
Humanity's Last Exam
3.7%4.7%
SciCode
20.9%21.2%
IFBench
31.5%31.8%
TerminalBench
2.3%3.8%
Granite 4.0 H Small2 wins
10 winsGemma 3 27B Instruct

Frequently Asked Questions

Which is cheaper, Granite 4.0 H Small or Gemma 3 27B Instruct?

Granite 4.0 H Small is cheaper overall. Its blended price (3:1 input/output ratio) is $0.11/M tokens vs $0.14/M for Gemma 3 27B Instruct.

Which model performs better on benchmarks?

Gemma 3 27B Instruct wins 10 out of 12 benchmarks compared to 2 for Granite 4.0 H Small. See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

Granite 4.0 H Small generates tokens faster at 265 tok/s vs 0 tok/s. However, Gemma 3 27B Instruct has lower time-to-first-token (0.00s vs 8.78s).

When should I use Granite 4.0 H Small vs Gemma 3 27B Instruct?

Choose based on your priorities: Granite 4.0 H Small for lower cost, Gemma 3 27B Instruct for stronger benchmark performance, and Granite 4.0 H Small for faster generation. For latency-sensitive apps, check the TTFT comparison above.