What does AI actually cost to ship a ticket?
- The problem
- Benchmark leaderboards are vanity metrics. They don't show what a model costs, or how it performs inside a real coding harness.
- What you get
- The exact cost of every harness against every model, run on the same repos and the same tickets.
- Why it matters
- Save money and time, and pick the best harness for your favorite model.
Benchmarks
Select a ticket to load the harness × model cost grid.
Cost matrix
Swipe to view more
| Model |  codex |  cursor |  devin |
|---|---|---|---|
| claude-fable-5 | — | ↑ $4.99 | — |
| claude-fable-5.1 | — | $4.37 | — |
| claude-haiku-5.5 | — | $0.07 | — |
| claude-opus-5 | — | $4.10 | — |
| claude-opus-5.5 | — | $1.59 | — |
| claude-sonnet-5 | — | $1.31 | — |
| claude-sonnet-5.5 | — | $0.62 | — |
| composer-2.5 | — | $0.30 | — |
| grok-4.6 | — | $1.28 | — |
| grok-4.7 | — | $2.01 | — |
| gpt-5.6-luna | $0.07 | $0.10 | $0.10 |
| gpt-5.6-sol | $1.98 | $1.92 | — |
| gpt-5.6-terra | $0.68 | $1.14 | $0.75 |
| gpt-6-astra | $2.41 | — | — |
| gpt-6-luna | ↓ $0.04 | — | — |
| gpt-6-sol | $0.71 | — | — |
| gpt-6.1-sol | $0.93 | — | — |
| swe-2 | — | — | $1.10 |
| deepseek-v4-flash | — | — | $0.08 |
| deepseek-v4-pro | — | — | $0.43 |
| deepseek-v4.1-flash | — | — | $0.08 |
| gemini-3.8-flash | — | $1.09 | — |
| glm-5.3 | — | — | $0.47 |
| kimi-k2.7-code | — | $0.45 | — |
| kimi-k3 | — | $1.08 | $1.29 |
| muse-spark-1.3 | — | $0.46 | — |
Swipe to view more
| Model | codex | cursor | devin |
|---|---|---|---|
| composer-2.5 | — | $0.68 | — |
| grok-4.6 | — | $5.51 | — |
| gpt-5.6-luna | ↓ $0.21 | — | — |
| gpt-5.6-terra | $1.09 | — | — |
| gpt-6-astra | $7.87 | — | — |
| swe-2 | — | — | ↑ $11.67 |
| gemini-3.8-flash | — | $2.99 | — |
| kimi-k2.7-code | — | $2.70 | — |
| muse-spark-1.3 | — | $0.83 | — |