Legacy long-context veteran versus the current mid generation.
gpt-4.1 · openai
gpt-5.4 · openai
Benchmark
| Workload Scenario | GPT-4.1 | GPT-5.4 | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $2.00 | $2.50 | −$0.50 |
| 1M output tokens (generation) | $8.00 | $15.00 | −$7.00 |
| 1M tokens · 70% input / 30% output mix | $3.80 | $6.25 | −$2.45 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.0114 | $0.0175 | −$0.0061 |
| Monthly scale (10K requests / day) | $3,420.00 | $5,250.00 | −$1,830.00 |
Negative difference = GPT-4.1 is cheaper. Positive = GPT-5.4 is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
GPT-4.1 is cheaper on both input and output rates, so it costs less at every input/output mix. Price alone still isn't the whole decision: capability, latency and context limits (GPT-4.1: 1M, GPT-5.4: 1.1M) may justify the premium for your task.
FAQ
On input, GPT-4.1 is cheaper ($2.00/M vs $2.50/M). On output, GPT-4.1 is cheaper ($8.00/M vs $15.00/M). The same model is cheaper on both sides, so it wins at every mix.
A chat-style request (4,000 input + 800 output tokens, 50% cached) costs $0.0114 on GPT-4.1 and $0.0175 on GPT-5.4 — GPT-4.1 is 1.5× cheaper for that workload.
GPT-4.1 supports 1,047,576 tokens (32.8K max output); GPT-5.4 supports 1,050,000 (128K max output). GPT-5.4 fits 1.0× more context, which matters for long documents and agents.