Tura
80% fewer tokens; 16.7 percentage points higher success.
Start new thread

Tura - 80% fewer tokens; 16.7 percentage points higher success.

by•
Across 348 long-horizon benchmark sessions, Tura used up to 80% fewer tokens on the rewrite benchmark and improved the DeepSWE pass rate by up to 16.7 percentage points compared with Codex CLI.

Add a comment

Replies

Best
Maker
📌
Tura is a local, open-source coding agent for developers who are tired of vague skill claims, token-saving extensions with no evidence, and agents without judgment wreck their repos. Across 20 DeepSWE v1.1 tasks, tested 60 sessions with GPT-5.6 SOL at High reasoning effort, Tura creates a substantial token-budget advantage by reducing repeated context and model round trips. You can spend that advantage in two ways. Direct turns most of it into lower cost: 83.5% fewer aggregate tokens than the official Codex CLI High configuration, with a verifier success rate of 65.0% versus 60.0%. Balanced puts more of the saved budget back into reasoning, investigation, and verification. It reached an 80.0% success rate—20 percentage points higher than Codex CLI High—while still using 49.6% fewer tokens

honestly the token savings caught me off guard, ran it on a gnarly refactor and it just chewed through way less than I'd expect for the quality it spit out

Maker

 Thank you. You can try some e2e frontend and full-stack tasks; normally, the diff is much larger than with refactoring.

the jump on DeepSWE without bloating token usage is genuinely impressive, most tools trade one for the other

Maker

 I'm fixing some timeout issues and will submit the full benchmark next week to deepswe.

Love that you're showing real benchmark numbers instead of vague promises. One thing I'd want to see: a way to compare token costs side-by-side against Codex or Cursor in the same session. A built-in cost diff readout would make it way easier to justify switching.

Saw the token savings on long rewrites and figured it was too good to be true, but the numbers held up in my own tests. Genuinely impressed by how much faster it gets through big refactors without losing track.