Frontier SWE
A benchmark for frontier software engineering that evaluates AI models on highly complex, open-ended engineering problems, with an emphasis on cutting-edge and open-ended engineering challenges.
Core Facts
Timeline (last 90 days)
On the Frontier SWE benchmark, Argon scores 55 and Opus 5.5 scores 62.3.
On Proximal's Frontier SWE benchmark (20 hours per task), Argon scored 55 and Astra 65.5.
Gemini 4 Argon significantly reduces the hedging expressions, repetition and boilerplate common to such models.
On Terminal-Bench 4.0, Claude Opus 5.5 leads with 66.4%, followed by GPT-6 Astra at 61.8% and Argon at 57.4%.
In 4 agentic coding tests, Argon won two, and ranked last in both the Frontier SWE and Terminal Bench tests.
All Facts (5)
Gemini 4 Argon significantly reduces the hedging expressions, repetition and boilerplate common to such models.
85%VerifiedOn Terminal-Bench 4.0, Claude Opus 5.5 leads with 66.4%, followed by GPT-6 Astra at 61.8% and Argon at 57.4%.
75%UnverifiedOn the Frontier SWE benchmark, Argon scores 55 and Opus 5.5 scores 62.3.
50%UnverifiedOn Proximal's Frontier SWE benchmark (20 hours per task), Argon scored 55 and Astra 65.5.
50%UnverifiedIn 4 agentic coding tests, Argon won two, and ranked last in both the Frontier SWE and Terminal Bench tests.
50%