The AI race hasn’t slowed down. In the last week, Google, Anthropic and Mistral AI all released new AI models ranging from widely accessible to super exclusive. Each focuses on different improvements, but the stakes go beyond benchmark bragging rights. We break down the latest releases, what’s new,and whether each one deserves your attention.
Some of the latest releases are incremental updates in model speed and price, where others are more substantial upgrades that push close to the frontier.
Below, we’ll break down some of the newest models that have been released and if and how you can get your hands on them yourself.
More from CNET
Claude Haiku 5.5 #
A major theme in new AI models lately is speed and cost efficiency, and that’s what Anthropic’s new Haiku 5.5 model brings to the table. It’s the first major update in a year for the company’s cheapest and fastest Claude. Anthropic says it’s designed for high-volume, cost-sensitive tasks and can reliably handle quick, repetitive workloads.
Compared with the previous Haiku model, Anthropic says its latest model costs about 75% less to run — an impressive price drop. It’s also seen substantial improvements across a series of benchmarks, including computer-use tasks, knowledge work, agentic coding and visual reasoning.
Haiku 5.5 is Anthropic’s first Haiku-level model to ship with an adjustable effort setting, allowing users to optimize the model for cost or intelligence. The latest model has seen significant improvements in alignment and cybersecurity, with stronger safeguards.
Claude Sonnet 5.5 #
One step up from Haiku, Anthropic’s new Claude Sonnet 5.5 model marks a clear upgrade over its 5.0 version. Like its peers, the latest Sonnet model is optimized for speed and affordability. Anthropic says there have been significant improvements to the model’s performance and collaboration capabilities.
Anthropic said Sonnet 5.5 has also seen enhancements in both alignment and cybersecurity capabilities, making it comparable to those of Opus 5, a previous iteration of its big sibling. The latest Sonnet model has also seen fairly sizable boosts in benchmarks for agentic coding and knowledge work, and Anthropic says that Sonnet 5.5’s performance boost is particularly noticeable in coding. Other improvements include better protection against distillation attacks, which are when attackers try to extract a model’s capabilities to build their own models without the built-in safeguards that Anthropic has implemented.
Gemini 4 Argon #
Last week, Google announced its latest frontier model, Gemini 4 Argon, but it’s currently limited to a small number of users. The model ships with a 1 million-token output limit and has enhanced deep reasoning capabilities, and allows you to take on bigger projects. The model is currently available only to trusted cybersecurity workers via Google’s Fairwind program. Competitors like Anthropic and OpenAI have done the same with major model updates, which will eventually trickle down to everyday users.
Like other Gemini models, Gemini 4 Argon is seeing enhancements in reasoning, coding and multimodal capabilities.
Mistral Large 4 #
Mistral AI’s latest flagship model, Mistral Large 4, dubbed “Le Chonk” due to its size, was released in a public preview earlier this week. According to Mistral AI, the new model ranks among the top open-weight models globally. It specializes in cybersecurity, geospatial and grounding capabilities and agentic AI. The public preview ends with an official rollout on Oct. 27. You can try the preview API here.