xAI, Anthropic, OpenAI, and Moonshot AI have all launched new advanced language models in July 2026, intensifying the competition in AI development. These releases mark a notable leap in AI capabilities, with improvements in token efficiency and longer task handling without human oversight.

xAI Launches Grok 4.5 with Competitive Pricing

On July 8, xAI introduced Grok 4.5, a model built on 1.5 trillion parameters. It leverages data collected from real developer interactions on Cursor, a coding platform acquired by SpaceXAI earlier this year. Grok 4.5 offers a 500,000 token context window and operates at $2 per million input tokens and $6 per million output tokens, undercutting competitors by over 60%. Elon Musk highlighted the model's speed and cost efficiency, and independent ratings placed Grok 4.5 fourth on the Artificial Analysis Intelligence Index, surpassing all open weight models. It also scored 83.3% on Terminal-Bench 2.1, which measures command-line engineering task performance. xAI claims Grok 4.5 requires only about a fifth of the output tokens compared to Opus 4.8 for similar tasks, dramatically lowering operational costs for extended AI sessions.

Anthropic Releases Claude Opus 5 as Default Model

Anthropic rolled out Claude Opus 5 on July 24, maintaining the same pricing structure as its predecessor Opus 4.8, with $5 per million input and $25 per million output tokens. Opus 5 features a 1 million token context window and supports an extended output of up to 128,000 tokens. The model includes an adjustable reasoning effort setting, introducing a new xhigh mode to tailor task complexity. While Opus 5 achieves new bests on Frontier-Bench and GDPval-AA evaluations for coding and knowledge tasks, it falls short of the specialized Claude Mythos 5 in cybersecurity. It is now the standard model for Claude Max subscriptions and the most powerful option on Claude Pro.

OpenAI and Moonshot AI Follow with Their Own Advances

OpenAI made the GPT-5.6 family widely available on July 9 after a preview period involving select vetted organizations. Moonshot AI also debuted Kimi K3, its largest model to date with 2.8 trillion parameters. These models share the focus on sustaining complex, multi-hour tasks from research through coding to structured reporting, aiming to reduce the need for human course correction during extended AI operations.

Markets responded to this flurry of AI model launches with increased interest but no immediate large-scale price swings.