Windsurf: Gemini 4 Argon Now Available in Devin Desktop and CLI at 75% Off

WindsurfOct 13, 2026View original changelog

Windsurf's Devin made Gemini 4 Argon available in Devin Desktop and Devin CLI, priced at 75% off list through October 27, 2026. Devin's proprietary FrontierCode 1.1 benchmark scores the model at 60.5%, the highest of any Gemini model on its leaderboard. Devin's evaluations describe a thorough model that reads widely, checks git history for repo conventions, and reviews its own diff before finishing.

Key Takeaways

Gemini 4 Argon is live in Devin Desktop and Devin CLI, giving users a new Google model option in both surfaces.
The model is offered at 75% off list price through October 27, 2026, combining Google's 50% introductory pricing with an extra 50% Devin discount.
Devin's FrontierCode 1.1 benchmark scores it at 60.5%, ahead of GPT-6.1 Sol (60.4%) and the top result for any Gemini model on that leaderboard.
It sits four points above Gemini 3.7 Flash, but still below Claude Sonnet 5.5, GPT-6 Astra and Claude Opus 5.5 on raw score.
Devin reports the model is thorough before it writes, reading widely, checking git history for conventions and reviewing its own diff.
It is strongest on bug fixes and refactors in backend and systems code, especially Python, C/C++ and PHP.

Gemini 4 Argon Arrives in Devin

Devin, the product Windsurf now ships under, added Gemini 4 Argon to both Devin Desktop and Devin CLI on October 13, 2026. Users can select the model alongside the other frontier models already offered in Devin, without any separate setup.

Launch Pricing: 75% Off List

Gemini 4 Argon is available at 75% off list price through October 27, 2026. The discount stacks two offers: Google's own 50% introductory pricing, which runs through December 1, 2026, and an extra 50% discount applied by Devin through October 27, 2026. After October 27, the Devin-side discount ends, while Google's introductory pricing continues until December 1.

How It Performs

On FrontierCode 1.1, Devin's proprietary benchmark that grades models on real-world engineering tasks for quality and mergeability, Gemini 4 Argon scores 60.5%. That edges past GPT-6.1 Sol (60.4%) and sits just behind GPT-6 Sol (60.7%). It is the highest-scoring Gemini model on the leaderboard, four points ahead of Gemini 3.7 Flash. Higher-scoring models on the same leaderboard include Claude Sonnet 5.5 (64.4%), GPT-6 Astra (64.5%) and Claude Opus 5.5 (65.3%), so Gemini 4 Argon lands in the middle of the pack on raw score while carrying a promotional price.

Working Style and Strengths

In Devin's evaluations, Gemini 4 Argon is thorough before it writes. It reads widely across a codebase, checks git history to pick up repository conventions, and reviews its own diff before finishing a task. Devin found it strongest on bug fixes and refactors in backend and systems code, particularly Python, C/C++ and PHP.

Why It Matters

For developers who already use Devin Desktop or the CLI, this adds a Google option that is competitively scored and, for the next two weeks, unusually cheap. The promotional window makes it a low-cost time to test Gemini 4 Argon on real backend and refactoring work and compare it with the models already in use.