Capability matrix
Raw benchmark parity rarely decides the pick — tooling, open weights, and provider reliability usually do.
Grok 4.7 leads on 2 of 2 shared benchmarks.
| Dimension | Grok 4.7 | Kimi K3 |
|---|---|---|
| GPQA Diamond | — | 92.9% |
| HLE | — | 43.5% |
| MathArena Apex | — | 65.6% |
| Terminal-Bench 2.1 | — | 88.3% |
| Terminal-Bench 3.0 | — | 17.7% |
| Terminal-Bench 4.0 | 38% | 12.6% |
| DeepSWE v1.1 | 71% | 67.5% |
| ProgramBench | — | 17.5% |
| NL2Repo-Bench | — | 58% |
| CyberGym | — | 80% |
| HLE (w/tools) | — | 59.8% |
| Automation-Bench | — | 46.7% |
| Agents' Last Exam | — | 27.6% |
| Chartography (w/tools) | — | 68.1% |
| BabyVision (w/tools) | — | 85.7% |
| ZeroBench-main (w/tools) | — | 41% |
| Blended price | $4.00 /1M | — |
| Full priceInput · Output · Cached input | $2.00 · $6.00 · $0.500 | — |
| Context window | 500K | — |
Adjust the monthly volume — every model recomputes instantly.
Estimates use list prices and ignore cached input, batch discounts, and off-peak rates.