Compare/Gemma 4 31B (Reasoning) vs gpt-oss-20B (low)

Gemma 4 31B (Reasoning)vsgpt-oss-20B (low)

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

Google

Gemma 4 31B (Reasoning)

Input
$0/M
Output
$0/M
Speed
TTFT
OpenAI

gpt-oss-20B (low)

Input
$0.06/M
Output
$0.2/M
Speed
266 tok/s
TTFT
0.48s

Winner by Category

Cheaper
Gemma 4 31B (Reasoning)
Faster (tok/s)
gpt-oss-20B (low)
Lower Latency
Gemma 4 31B (Reasoning)
Benchmarks (3-8)
gpt-oss-20B (low)

Pricing Comparison

MetricGemma 4 31B (Reasoning)gpt-oss-20B (low)
Input ($/M tokens)$0$0.06
Output ($/M tokens)$0$0.2
Cost for 1M input + 100K output tokens:
Gemma 4 31B (Reasoning)$0.00
gpt-oss-20B (low)$0.08

Speed Comparison

Output Speed (tokens/s) — higher is better
Gemma 4 31B (Reasoning)
gpt-oss-20B (low)
266 tok/s
Time to First Token (seconds) — lower is better
Gemma 4 31B (Reasoning)
gpt-oss-20B (low)
0.48s

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
20.8
Coding Index
14.4
Math Index
62.3
GPQA Diamond
85.7%61.1%
MMLU-Pro
71.8%
LiveCodeBench
65.2%
AIME 2025
62.3%
MATH-500
Humanity's Last Exam
22.7%5.1%
SciCode
34.0%
IFBench
75.6%57.8%
TerminalBench
4.5%
Gemma 4 31B (Reasoning)3 wins
8 winsgpt-oss-20B (low)

Frequently Asked Questions

Which is cheaper, Gemma 4 31B (Reasoning) or gpt-oss-20B (low)?

Gemma 4 31B (Reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.00/M tokens vs $0.09/M for gpt-oss-20B (low).

Which model performs better on benchmarks?

gpt-oss-20B (low) wins 8 out of 12 benchmarks compared to 3 for Gemma 4 31B (Reasoning). See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

gpt-oss-20B (low) generates tokens faster at 266 tok/s vs 0 tok/s. Gemma 4 31B (Reasoning) also has lower time-to-first-token (0.00s vs 0.48s).

When should I use Gemma 4 31B (Reasoning) vs gpt-oss-20B (low)?

Choose based on your priorities: Gemma 4 31B (Reasoning) for lower cost, gpt-oss-20B (low) for stronger benchmark performance, and gpt-oss-20B (low) for faster generation. For latency-sensitive apps, check the TTFT comparison above.