Compare/Tri-21B-think Preview vs GPT-5.4 nano (medium)

Tri-21B-think PreviewvsGPT-5.4 nano (medium)

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

Trillion Labs

Tri-21B-think Preview

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

GPT-5.4 nano (medium)

Input
$0.2/M
Output
$1.25/M
Speed
171 tok/s
TTFT
2.38s

Winner by Category

Cheaper
Tri-21B-think Preview
Faster (tok/s)
GPT-5.4 nano (medium)
Lower Latency
Tri-21B-think Preview
Benchmarks (0-7)
GPT-5.4 nano (medium)

Pricing Comparison

MetricTri-21B-think PreviewGPT-5.4 nano (medium)
Input ($/M tokens)$0$0.2
Output ($/M tokens)$0$1.25
Cost for 1M input + 100K output tokens:
Tri-21B-think Preview$0.00
GPT-5.4 nano (medium)$0.33

Speed Comparison

Output Speed (tokens/s) — higher is better
Tri-21B-think Preview
GPT-5.4 nano (medium)
171 tok/s
Time to First Token (seconds) — lower is better
Tri-21B-think Preview
GPT-5.4 nano (medium)
2.38s

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
20.038.1
Coding Index
7.435.0
Math Index
GPQA Diamond
53.8%76.1%
MMLU-Pro
LiveCodeBench
AIME 2025
MATH-500
Humanity's Last Exam
5.7%14.7%
SciCode
17.8%38.4%
IFBench
47.1%64.4%
TerminalBench
2.3%33.3%
Tri-21B-think Preview0 wins
7 winsGPT-5.4 nano (medium)

Frequently Asked Questions

Which is cheaper, Tri-21B-think Preview or GPT-5.4 nano (medium)?

Tri-21B-think Preview is cheaper overall. Its blended price (3:1 input/output ratio) is $0.00/M tokens vs $0.46/M for GPT-5.4 nano (medium).

Which model performs better on benchmarks?

GPT-5.4 nano (medium) wins 7 out of 12 benchmarks compared to 0 for Tri-21B-think Preview. See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

GPT-5.4 nano (medium) generates tokens faster at 171 tok/s vs 0 tok/s. Tri-21B-think Preview also has lower time-to-first-token (0.00s vs 2.38s).

When should I use Tri-21B-think Preview vs GPT-5.4 nano (medium)?

Choose based on your priorities: Tri-21B-think Preview for lower cost, GPT-5.4 nano (medium) for stronger benchmark performance, and GPT-5.4 nano (medium) for faster generation. For latency-sensitive apps, check the TTFT comparison above.