Compare/GPT-5.4 nano (xhigh) vs Claude 3 Haiku

GPT-5.4 nano (xhigh)vsClaude 3 Haiku

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

OpenAI

GPT-5.4 nano (xhigh)

Input
$0.2/M
Output
$1.25/M
Speed
178 tok/s
TTFT
2.70s
Anthropic

Claude 3 Haiku

Input
$0.25/M
Output
$1.25/M
Speed
131 tok/s
TTFT
0.47s

Winner by Category

Cheaper
GPT-5.4 nano (xhigh)
Faster (tok/s)
GPT-5.4 nano (xhigh)
Lower Latency
Claude 3 Haiku
Benchmarks (7-2)
GPT-5.4 nano (xhigh)

Pricing Comparison

MetricGPT-5.4 nano (xhigh)Claude 3 Haiku
Input ($/M tokens)$0.2$0.25
Output ($/M tokens)$1.25$1.25
Cost for 1M input + 100K output tokens:
GPT-5.4 nano (xhigh)$0.33
Claude 3 Haiku$0.38

Speed Comparison

Output Speed (tokens/s) — higher is better
GPT-5.4 nano (xhigh)
178 tok/s
Claude 3 Haiku
131 tok/s
Time to First Token (seconds) — lower is better
GPT-5.4 nano (xhigh)
2.70s
Claude 3 Haiku
0.47s

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
44.412.3
Coding Index
43.96.7
Math Index
GPQA Diamond
81.7%37.4%
MMLU-Pro
LiveCodeBench
15.4%
AIME 2025
MATH-500
39.4%
Humanity's Last Exam
26.5%3.9%
SciCode
46.9%18.6%
IFBench
75.9%36.1%
TerminalBench
42.4%0.8%
GPT-5.4 nano (xhigh)7 wins
2 winsClaude 3 Haiku

Frequently Asked Questions

Which is cheaper, GPT-5.4 nano (xhigh) or Claude 3 Haiku?

GPT-5.4 nano (xhigh) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.46/M tokens vs $0.50/M for Claude 3 Haiku.

Which model performs better on benchmarks?

GPT-5.4 nano (xhigh) wins 7 out of 12 benchmarks compared to 2 for Claude 3 Haiku. See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

GPT-5.4 nano (xhigh) generates tokens faster at 178 tok/s vs 131 tok/s. However, Claude 3 Haiku has lower time-to-first-token (0.47s vs 2.70s).

When should I use GPT-5.4 nano (xhigh) vs Claude 3 Haiku?

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