OpenAI released GPT-5.2, a refined iteration of GPT-5 with improvements in multi-step reasoning, code generation accuracy, and instruction following. It joins GPT-5 Pro, Mini, and Nano as the most capable model in the GPT-5 family.

Claude Opus 4.6 pushes Anthropic's frontier further with extended context handling and improved agentic capabilities. Claude Sonnet 4.5 and Haiku 4.5 round out the Claude 4 family, offering speed and cost trade-offs for different use cases.

xAI released Grok 4, a major generation leap over Grok 3, with Grok 4.1 Fast Reasoning as a specialised variant optimised for chain-of-thought and mathematical tasks at low latency.

Google DeepMind announced the Gemini 3 family. Gemini 3 Pro delivers state-of-the-art multimodal performance while Gemini 3 Flash targets production workloads with fast inference and competitive quality.

Mistral AI updated their mid-range lineup. Mistral Medium 3.1 improves instruction following and function calling, while Small 3.2 (a 14B parameter model) offers an open-weight option competitive with larger proprietary models.

DeepSeek released version 3.2 with two variants: Chat for conversational and instruction tasks, and Reasoning for extended chain-of-thought problems. Both are available via API and as open-weight models.

OpenAI introduced GPT-4.1 alongside two smaller variants. GPT-4.1 Mini and Nano bring GPT-4 level quality to high-throughput and latency-sensitive applications at significantly reduced cost.

Google expanded the Gemini 2.5 series with Gemini 2.5 Pro offering deeper reasoning and a one-million-token context window, and Gemini 2.5 Flash Lite as an ultra-fast, low-cost option.

Anthropic launched the Claude 4 series, marking a significant generational upgrade. Opus 4.0 and Sonnet 4.0 were the initial releases, with Opus 4.1 following as an improved iteration with better tool use and code generation.

Mistral AI's flagship Mistral Large 3 arrived with enhanced multilingual capabilities, stronger coding performance, and a larger context window, maintaining Mistral's focus on efficient European-developed AI.

xAI's Grok 3 debuted with real-time web access and strong performance on reasoning benchmarks. Grok 3 Mini offers a lighter, faster version suited for everyday tasks.

Google released the Gemini 2.0 generation starting with Flash variants. Gemini 2.0 Flash brought multimodal live streaming capabilities, while Flash Lite targeted mobile and edge deployments.