Moonshot AI Launches Kimi 3 With Limited Launch Information
The Chinese firm Moonshot AI has introduced its Kimi 3 model as the country's largest AI system, yet the launch comes with no supporting data on performance or usage terms.
Frontier AI Models
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.
OpenAI GPT-5.6 Models Launch on Amazon Bedrock with Matched Pricing
The Sol flagship, Terra balanced, and Luna efficient models extend agentic coding and reasoning to AWS customers through Bedrock's inference engine while aligning costs with existing commitments.
Anthropic Redeploys Claude Fable 5 Globally After Export Controls Lifted
The model returns with safeguards that route select queries to Opus 4.8 while Sonnet 5 and Opus 4.8 advance agentic performance toward GPT-5.5 levels at competitive pricing.
Anthropic Releases Claude Sonnet 5 and Restores Fable 5 for Agentic AI Access
Dual updates position mid-tier models closer to Opus 4.8 performance while addressing export control issues through improved safeguards and pricing.
OpenAI Reorganizes Safety into Research Division After Heidecke Exit and GPT-5.6 Release
The integration of safety teams under research leadership follows multiple executive departures and the launch of a model family with increased misalignment rates, prompting questions about the future of independent safety oversight at the company.
Google's Gemini 3.5 Flash Delivers Frontier Agentic Performance at Flash Speed
The May 2026 release positions the model as default for the Gemini app and AI Mode in Search, prioritizing efficient execution of complex tasks over larger model sizes favored by some rivals.
Empero AI Releases Qwythos-9B-Claude-Mythos-5-1M Open-Weight Model
The June 19, 2026 launch introduces a 9B parameter model with 1M token context, native tool-calling and self-correction, built through post-training on Claude traces from an uncensored Qwen base.
SpaceXAI Grok 4.5 Delivers Opus-Class Coding at Disruptive Pricing
The July 2026 release targets developers with a 500,000 token context window, strong benchmark scores on coding tasks, and pricing that undercuts leading alternatives while maintaining high efficiency.
OpenAI GPT-5.6 Sol Tops CritPt Physics Benchmark
The latest model variant from OpenAI has secured the leading position on a benchmark designed to test reasoning on genuine research-level physics problems.
Dolphin Mistral 24B Venice Edition Integrates on OpenRouter with 128k Context
The platform addition supplies a fine-tuned uncensored 24B model from Mistral-Small-24B-Instruct-2501 through paid and free tiers, developed via dphn.ai and Venice.ai collaboration to emphasize user steerability.
Meta Releases Muse Spark 1.1 Agentic Coding Model via Public API
The upgrade positions Meta as a competitor in agentic coding with an 80 percent benchmark score and million-token context, available through a new developer API and consumer app.
OpenAI GPT-Live Introduces Full-Duplex Voice AI for Natural Conversations
The July 8, 2026 launch equips ChatGPT Voice with models that listen and speak simultaneously, supporting natural speech elements and complex task delegation to improve user interactions across platforms.
Grok 4.5 Launches Claiming Opus 4.7 Parity at 80 TPS and $2/$6 Pricing
SpaceXAI partners with Cursor to release Grok 4.5, a model designed for coding and agentic tasks that aims to deliver Opus 4.7 level performance at reduced costs and increased inference speeds for developers and knowledge workers.
SpaceXAI Releases Grok 4.5: Opus-Class Coding Model at $2 per Million Tokens
The new release from SpaceXAI delivers frontier-level performance for coding and agentic tasks through joint training with Cursor, combining high intelligence with notable advantages in speed, token efficiency, and pricing.
Grok 4.5 Launches with Focus on Efficiency and Agentic Coding
xAI's Grok 4.5 model enters the frontier AI space emphasizing lower costs and token efficiency while achieving performance levels comparable to GPT-5.5 in coding benchmarks.
Grok 4.5 Launches as SpaceXAI Frontier Model with Cursor Partnership
The jointly developed system targets coding and agent workloads through Colossus training while delivering leading efficiency metrics and benchmark results ahead of broader EU availability.
xAI Readies Grok 4.5 for Public Launch as Opus-Class Challenger
The model enters the market on July 9 after beta validation at SpaceX and Tesla and uses a 1.5T V9 base with Cursor training to deliver efficiency gains against established competitors.
Agents-A1 Matches Trillion-Scale Performance with 35B Parameters via Horizon Scaling
InternScience at Shanghai Artificial Intelligence Laboratory open-sources a quantized 35B MoE model trained on extended agent trajectories to rival larger systems on benchmarks like SEAL-0 and IFBench.
Claude Fable 5 Reclaims SWE-bench Pro Top Spot at 80.3 Percent
Anthropic's model demonstrates superior results on complex coding benchmarks and delivers measurable efficiency in enterprise-scale Ruby codebase migrations validated by Stripe.
Claude Fable 5 Tops Senior SWE-bench Leaderboard at 27.9%
Anthropic's model advances to the front of senior-level agent evaluations with a clear margin over prior versions, supported by full reproducibility through the Harbor Hub platform.
Frequently asked about Frontier AI Models
What counts as a frontier AI model?
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.
How often are new frontier models released?
Major labs ship a flagship roughly every 6–12 months, with point upgrades in between. This hub updates on every confirmed release.
Where can I track frontier-model benchmarks?
Each article here cites the primary benchmark source — the lab model card or an independent eval — so you can verify the numbers directly.