Halv

What it is

Halv is a desktop application that lets developers run Claude Code, Codex, Kimi, and GLM models in a single workspace. The desktop app is free and open source under the Apache 2.0 license, available for macOS, Windows, and Linux.

Halv Cloud is a paid service that adds JEV agent routing, which automatically selects the most cost-effective model and reasoning effort for each task. The routing is included with Halv Cloud at no additional fee. A benchmark on 42 paired runs showed Halv recorded 57.1% lower model costs compared to vanilla Codex while achieving the same number of verifier passes.

The application includes features like a synced savings dashboard, history across machines, automatic account rotation, and local tools for code navigation, command output filtering, and model context processing. Halv Cloud costs US$10 per month with the option to cancel anytime.

Questions from the bench

Did Halv solve more tasks?

Both Halv and vanilla Codex passed 25 of 42 runs. They did not solve the same set of repetitions. Halv gained passes on ArcadeDB-4281 and HugeGraph-3037, but lost two Perry passes.

Did Halv use fewer tokens?

No. Halv used 170,114,775 total tokens versus 48,755,656 for vanilla Codex. The routed workflow recorded lower model cost while using 3.49 times as many tokens.

Did both arms use the same model?

Both coordinators used Astra with medium reasoning. Vanilla Codex worked alone. Halv added Luna and Sol workers with different reasoning efforts, plus Crux, RTK, and Headroom.

What does the 57.1% savings measure?

It compares recorded model cost across 84 selected runs: 42 Halv runs and 42 vanilla runs. Costs include coordinator sessions, worker sessions, and valid runs that failed verification. They exclude infrastructure-invalid attempts and the rolled-back policy experiment.

Does this reduce my subscription price by 57.1%?

No. These are recorded token-based cost estimates for the selected benchmark runs. They are not invoices or a reduction in fixed subscription prices.