Frontend benchmarks
Frontend Code Arena
Illustrative snapshot from arena.ai Code Arena | WebDev (Elo from blind human votes on frontend tasks, Jul 16 2026). X-axis is blended public API $/MTok — preference Elo has no $/task. Not live OpenCodex metering.
Costs on this board are estimates or API-price blends — score/$ ranking and Best value are disabled until every row has a measured cost-per-task from the source.
| Model | Effort | Score | Cost / task | Use case |
|---|---|---|---|---|
| kimi-k3 | — | 1679 ±17 | $9 | Frontier, Workhorse |
| claude-fable-5 | — | 1631 ±13 | $30 | Planner, Frontier |
| gpt-5.6-sol | xhigh | 1618 ±13 | $17.5 | Frontier, Planner |
| glm-5.2 | max | 1587 ±10 | $2.9 | Workhorse, Cheap subagent |
| claude-opus-4.8 | thinking | 1562 ±9 | $15 | Frontier, Planner |
| grok-4.5 | — | 1558 ±13 | $4 | Workhorse, Frontier |
| claude-opus-4.7 | thinking | 1558 ±7 | $15 | Frontier |
| claude-opus-4.7 | — | 1555 ±7 | $15 | Frontier |
| claude-opus-4.6 | thinking | 1542 ±6 | $15 | Workhorse |
| claude-sonnet-5 | high | 1542 ±12 | $6 | Workhorse |
| claude-opus-4.6 | — | 1536 ±6 | $15 | Workhorse |
| claude-opus-4.8 | — | 1534 ±9 | $15 | Frontier |
| glm-5.1 | — | 1526 ±9 | $2.9 | Workhorse, Cheap subagent |
| claude-sonnet-4.6 | — | 1522 ±6 | $9 | Workhorse |
| kimi-k2.6 | — | 1515 | $2.13 | Cheap subagent, Workhorse |
| gpt-5.5 | xhigh | 1504 ±7 | $17.5 | Frontier, Workhorse |
| gpt-5.5 | high | 1482 ±7 | $17.5 | Workhorse |
| gpt-5.4 | high | 1457 ±17 | $8.75 | Workhorse |