FlashNEWS BRIEFING

Cognition Evaluation Shows Gemini 4 Neon Matches Claude Opus 5.5 on Code at One-Third the Cost

1 min read

88Z AI, 20 November — Cognition evaluated Gemini 4 Neon on its FrontierCode 2.0 benchmark, finding it scores within one point of Claude Opus 5.5 at approximately one-third the cost. As the strongest model in the Gemini series with a per-task cost of $4.38, it is now available for use in Devin Desktop and Devin CLI.

At a glance

  • Gemini 4 Neon scores 46.2% on FrontierCode 2.0, trailing Claude Opus 5.5 by only one point
  • Costs $4.38 per task, approximately one-third the price of its competitor
  • Now available for immediate use in Devin Desktop and Devin CLI
Chart comparing Gemini 4 Neon and Claude Opus 5.5 performance on FrontierCode 2.0 by Cognition
Chart comparing Gemini 4 Neon and Claude Opus 5.5 performance on FrontierCode 2.0 by Cognition

Benchmark Performance and Cost Efficiency

Cognition officially released evaluation results for Gemini 4 Neon. On its FrontierCode 2.0 benchmark, which measures whether a pull request should be merged, Gemini 4 Neon scored within one point of Claude Opus 5.5. Despite nearing top-tier performance, it offers significant cost efficiency, priced at approximately one-third the cost of Opus 5.5.

Model Standing and Practical Availability

Data indicates that Gemini 4 Neon is the strongest model in the Gemini series, achieving a score of 46.2% on the benchmark with a cost of $4.38 per task. The model is now immediately available for developers to use within Devin Desktop and Devin CLI for code assistance workflows.

Sources & editorial note

This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.

  1. @cognitionSource published

Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.

Related reading

All stories
88Z.ai