Mercury 2 vs Mercury 2.5
Different efforts: high · thinking
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
Mercury 217.1
Mercury 2.511.8
Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite
Mercury 20
Mercury 2.50
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
Mercury 21.7
Mercury 2.52.4
Arena Text · creative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool
Mercury 21,293
Mercury 2.5—
| Benchmark | Mercury 2 | Mercury 2.5 |
|---|---|---|
| Humanity's Last Exam | 17.1 | 11.8 |
| Terminal-Bench 4.0 | 0 | 0 |
| AutomationBench-AA | 1.7 | 2.4 |
| Arena Text · creative writing | 1,293 | — |
Specs
| Spec | Mercury 2 | Mercury 2.5 |
|---|---|---|
| Input, per 1M | $0.25 | $0.20 |
| Output, per 1M | $0.75 | $0.75 |
| Cost per task? | $0.039 | $0.12 |
| Speed, tok/s | 487 | 781 |
| Context | 128K | 260K |
| Weights | Closed | Closed |
| Released | Feb 2026 | Sep 2026 |
| See images | — | — |
| Read PDFs | — | — |
| Hear audio | — | — |
| Watch video | — | — |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
Mercury 2 → Mercury 2.5
Nothing known.
Mercury 2.5 → Mercury 2