#FrontierCode Articles

5 articles

AI Rating: High AI Rating: Medium AI Rating: Low

Chat & General Models Image Generation Video Generation Voice & Music Coding Search & Agents

Devin News 09/29/2026 Read AI Rating: Medium

Devin Adds ChatGPT Sign-In and Support for OpenAI's GPT-6.1 Sol

Cognition rolled out two OpenAI-linked updates to Devin on September 29, 2026: a ChatGPT sign-in that lets eligible Plus/Pro users draw OpenAI model usage from their existing plan, and day-one support for GPT-6.1 Sol, which matches prior Sol models' benchmark scores at a lower price.

Devin News 09/28/2026 Read AI Rating: Medium

Claude Sonnet 5.5 Comes to Devin, Beating Fable 5.1 on FrontierCode at a Lower Cost

Cognition added Claude Sonnet 5.5 to Devin on September 28, 2026, reporting a 64.4% score on the FrontierCode 1.1 benchmark, more than 30% faster output generation, and up to 30% lower cost per task than Sonnet 5, at unchanged token pricing.

Devin News 09/28/2026 Read AI Rating: Medium

Devin Gets Up to 40% Cheaper Across Fusion, Normal, Ultra, and Devin Review

Cognition cut Devin's pricing across the board on September 28, 2026, with Fusion and Normal modes 30-40% cheaper, Ultra 15-20% cheaper, and Devin Review up to 70% cheaper, driven by newer models and harness-level efficiency work.

Devin News 07/13/2026 Read AI Rating: Low

Cognition Says Pairing Fable with Its "Fusion" Sidekick Setup Beats Opus on Cost and Score

Devin maker Cognition has published results showing that combining its multi-agent "Fusion" setup with Anthropic's Fable 5 costs less and scores higher than running Opus 4.8, based on 3,000 evaluation sessions.

Devin News 07/07/2026 Read AI Rating: Low

Cognition Releases FrontierCode 1.1, an Updated Coding Evaluation Benchmark

Devin maker Cognition has updated its internal code-quality benchmark to FrontierCode 1.1, refining how legitimate internet use is separated from unfair shortcuts, relaxing overly strict grading criteria, and publishing new scores for Sonnet 5 and Fable 5.