Releases, benchmarks and capability jumps from OpenAI, Anthropic, Google DeepMind, Meta and xAI.
Frontier models are the largest, most capable general-purpose AI systems at the edge of what is technically possible — today that means GPT-class models from OpenAI, Anthropic's Claude, Google DeepMind's Gemini, Meta's Llama and xAI's Grok. This section tracks every major release, benchmark result and capability shift across the leading labs, with sourced analysis of what each change means for the people who build on these models.
The 1.05 trillion-parameter open-weight model, trained in European datacenters, ties GLM-5.3 on a leading coding-agent index while using fewer tokens and now runs on Open Math OS.
The 753B-parameter model, released Nov. 2, 2026 under an MIT license, self-reports a five-point edge over GPT-6-Astra on FrontierScience-Research and enters a market that includes Meituan's LongCat-2.5 agent.
The firm's OpenAI-powered Debrief and AskResearchGPT tools translate meeting notes and 70,000 research reports into advisor productivity gains.
The six-model LFM3 family runs text, vision, audio, omni, and decision workloads on a shared 1.7B backbone, targeting phones, laptops, and edge hardware over cloud-scale deployment.
Google's next frontier model leads the updated Artificial Analysis Intelligence Index at 59 and posts record cybersecurity benchmark scores, but initial access is limited to vetted defenders under the Fairwind Program.
Google DeepMind's new world model generates synchronized audio and interactive dialogue alongside visuals, lets users steer scenes in natural language, and doubles the consistency window of Genie 3 — raising the stakes in the race with Runway over interactive simulation.
Artificial Analysis' new code-to-video benchmark reveals a sharp quality-versus-cost split between frontier models and open-weights systems.
The Oct. 26 release pairs a 1.05-million-token context window and computer use with API pricing that is 50% below GPT-6.1 Sol.
The company says its internal models produced landmark proofs in complexity theory, harmonic analysis and quantum field theory. The original proposers of ETH caution that involvement is not certification.
The French lab pairs state-of-the-art transcription with a real-time voice agent SDK, aiming to make EU data residency a selling point against US rivals.
The 42B-active-parameter model scores 39 on the Artificial Analysis Intelligence Index and runs locally on dual Mac Studios, but the release raises safety and competitive-positioning questions.
The Apache 2.0-licensed model matches GPT-6.1 Sol on five DNA benchmarks while cutting inference costs by roughly 1,800x, according to Hugging Face.
The Paris-based model maker enters the real-time voice stack with 50-plus-language transcription, $0.0035-per-minute pricing and EU data-control options.
How a British universal bank built a governed, production-grade AI operation on Anthropic's Claude, with usage metrics peers can benchmark.
Nvidia's Nemotron 3.5 family spans a 120B vision-language flagship for always-on agents and two edge-optimized Bolt models, extending open-weight agentic AI from data centers to RTX desktops and Jetson devices.
The ChatGPT maker is compressing its fundraising timeline to lock in capital before a rival's public debut, with SoftBank and MGX in talks and compute expansion the stated priority.
The October rollout supplies frontier agentic coding and visual performance through US infrastructure that applies zero data retention while open weights follow four days later.
Google DeepMind's frontier model targets long-horizon software engineering, enterprise knowledge work, and cyber defense, with an introductory API price of $2 per million input tokens.
The Oct. 17 refresh of DeepSeek's open-weight Flash model posts record DeepSWE and Automation-Bench scores, cuts output overhead and adds Search and Fetch APIs without changing the price sheet.
Open-weight release from Nvidia and the Nemotron Coalition gives enterprises post-training control and single-GPU throughput that rivals larger closed reasoning systems.
A frontier model is a general-purpose AI system trained at the largest compute scales — models like GPT-class, Claude, Gemini, Llama and Grok — whose capabilities exceed prior systems on broad benchmarks.
Major labs ship a flagship roughly every 6–12 months, with point upgrades in between. This hub updates on every confirmed release.
Each article here cites the primary benchmark source — the lab model card or an independent eval — so you can verify the numbers directly.
Recommended reading
The field-tested playbook for leading AI transformation — from literacy and governance to ROI, security, and scale.
Get the book on Amazon (opens in a new tab)Paid link. As an Amazon Associate, we earn from qualifying purchases.