Verdict
computed from the published numbersGPT-5.4 wins on Terminal-Bench, on intelligence and on context window. Mercury 2 wins on price.
Price
Mercury 2
Terminal-Bench
GPT-5.4
Intelligence
GPT-5.4
Context window
GPT-5.4
Key takeaways
who wins on what, from the published numbersGPT-5.4 wins
- Larger context window (1.1M vs 128K)
- Higher max output (128K vs 50K)
- More serving providers (6 vs 3)
- Supports vision
- Has reasoning mode
- Has web search
Mercury 2 wins
- Cheaper input ($2.50 vs $0.25 /1M)
- Cheaper output ($15.00 vs $0.75 /1M)
Cheaper overall
Mercury 2
$0.25 + $0.75 /1M
Larger context
GPT-5.4
1.1M
More providers
GPT-5.4
6 providers
Pricing & specs
cheapest listed route · per 1M tokensOn-demand listed rates, OpenRouter routed quotes and Vercel reseller quotes. Enterprise commitments can be materially lower.
| GPT-5.4 | Mercury 2 | |
|---|---|---|
| Input $/1M | $2.50via Databricks | $0.25via Inception best |
| Output $/1M | $15.00via OpenAI | $0.75via Inception best |
| Context window | 1.1M best | 128K |
| Max output | 128K best | 50K |
| Serving providers | 6 best | 3 |
| Date | Released Mar 5, 2026 | Released Feb 24, 2026 |
Benchmarks
Benchmark scores, side by side| GPT-5.4 | Mercury 2 | |
|---|---|---|
| Intelligence | 39.0 best | 13.8 |
| Terminal-Bench | 78% best | 27% |
| GPQA Diamond | 92% best | 77% |
| Output speed | 92 t/s | 554 t/s best |
| Latency (TTFT) | 192.30s | 6.83s best |
Cost calculator
choose a serving route for each sideInput tokens / month
Output tokens / month
GPT-5.4
via Databricks
$27
Mercury 2
via Inception
$2.00
Capabilities
feature support, side by side| Feature | GPT-5.4 | Mercury 2 |
|---|---|---|
| Vision | ✓ | — |
| Tool use | ✓ | ✓ |
| Reasoning | ✓ | — |
| Audio | — | — |
| Prompt caching | ✓ | ✓ |
| Structured output | ✓ | ✓ |
| Web search | ✓ | — |
Related comparisons
Frequently asked questions
Which is cheaper, GPT-5.4 or Mercury 2?
Mercury 2 is cheaper for input ($2.50 vs $0.25 per 1M tokens) from Inception. Mercury 2 is cheaper for output ($15.00 vs $0.75 per 1M tokens) from Inception.
What is the context window difference between GPT-5.4 and Mercury 2?
GPT-5.4 has a 1.1M-token context window and Mercury 2 has 128K tokens, so GPT-5.4 can take a larger prompt.
How do GPT-5.4 and Mercury 2 compare on benchmarks?
On the tracked benchmarks, GPT-5.4 leads on Terminal-Bench, and GPT-5.4 leads on intelligence.
Listed prices from LiteLLM, OpenRouter and Vercel · last checked Oct 14, 2026
Every weekday
AI moves fast. Here's your debrief.
News, analysis, tools, and more.
For people who build with AI