Anthropic Launches Sonnet 5.5, Promises 30% Speed Increase

Anthropic unveils Sonnet 5.5, touted as faster and cheaper than Sonnet 5. The company highlights advancements in agentic programming and cyber capabilities comparable to Opus 5, with protections aligned with its flagship models. A new Haiku is set to be announced in the coming weeks, amid closely timed announcements from OpenAI and Meta.
Anthropic aligns Sonnet 5.5 protections with Fable and Opus
Anthropic states that Sonnet 5.5 is the first model in the Sonnet range to be subjected to the same cyber protections as Fable and Opus. The company also claims that Sonnet 5.5 has cyber capabilities comparable to those of Opus 5. According to Anthropic's benchmarks, Sonnet 5.5 outperforms Opus 5.5 in agentic programming. Within the company's lineup, Sonnet remains less powerful than Opus but may be preferred in certain cases for its agility.
Announced performance: +30% speed and reduced token expenditure
Anthropic announces that Sonnet 5.5 exhibits a speed increase of 30% compared to Sonnet 5 and significantly reduces token consumption. The model is described as noticeably faster and more economical than its predecessor, while being designed to assist in everyday tasks such as programming or drafting office documents. Sonnet 5 was introduced about three months ago, with a focus on the efficiency of agentic deployment, meaning the ability to run agents at a lower cost than other players in the industry. The current advantage observed in the field of agentic programming likely stems from the ability to generate multiple agents without exceeding cost thresholds.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
A new Haiku on the way and very active competition
Anthropic plans to launch a new version of Haiku, its smallest model, in the coming weeks, without providing a specific date. The previous week, OpenAI unveiled various new models, including improved iterations of Sol and Luna, which correspond to its mid-range and lower-cost models. Meta, for its part, announced a new model, presented as set to power a future feature related to its smart glasses. These announcements come in a context where major AI labs are ramping up model launches.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.