{"id":41,"date":"2026-01-14T11:00:00","date_gmt":"2026-01-14T11:00:00","guid":{"rendered":"https:\/\/numriq.com\/llm-recap-q4-2025\/"},"modified":"2026-01-14T11:00:00","modified_gmt":"2026-01-14T11:00:00","slug":"llm-recap-q4-2025","status":"publish","type":"post","link":"https:\/\/numriq.com\/en\/llm-recap-q4-2025\/","title":{"rendered":"Anthropic, OpenAI, Mistral: what actually shipped in Q4 2025"},"content":{"rendered":"

October to December 2025 was the busiest AI model release quarter since the market began. Here’s what we kept, filtering marketing from real change for businesses.<\/p>\n

Anthropic: Claude 4.6 and the arrival of agents<\/h2>\n

Claude 4.6 Opus shipped in November, with clear improvements on multi-step reasoning and very long context handling (1M tokens). Quieter but more important: Claude for Agents, a variant optimized for agentic workflows with better tool control and fewer infinite loops.<\/p>\n

For businesses: Sonnet remains the sweet spot for most tasks. Opus for complex analysis. Claude for Agents for multi-step production workflows.<\/p>\n

OpenAI: GPT-5 and the price war<\/h2>\n

GPT-5 launched in October. Impressive capabilities, especially in reasoning and code. But the most strategic move was the 60% price cut on GPT-4o and GPT-4o-mini in November. It’s now the cheapest option for high-volume tasks.<\/p>\n

Mistral: Large 3 and the European angle<\/h2>\n

Mistral Large 3 shipped in December. Performance close to Claude Sonnet on French and European languages. European hosting, natively GDPR-compliant. Interesting for businesses with Law 25 constraints and multi-jurisdiction operations.<\/p>\n

Google: Gemini 2 Pro<\/h2>\n

Gemini 2 Pro launched in November. Excellent on multimodal (video, audio, images). If your use case involves non-textual media, it’s the first option to test.<\/p>\n

Our recommendation by use case<\/h2>\n

For a French-speaking B2C chatbot: Claude Sonnet, hosted on AWS Canada.<\/p>\n

For legal or financial document analysis: Claude Opus.<\/p>\n

For high-volume classification (millions of calls\/month): GPT-4o-mini.<\/p>\n

For multimedia processing: Gemini 2 Pro.<\/p>\n

For a business with European operations or strict compliance: Mistral Large 3.<\/p>\n

The rule hasn’t changed: no good model in absolute terms, just a good model for your case. But the bench has widened, and the price-performance ratio has clearly improved across the board.<\/p>\n","protected":false},"excerpt":{"rendered":"

Late 2025 was packed with model releases. Here’s what actually changes for businesses, past the announcement noise.<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[3],"tags":[],"class_list":["post-41","post","type-post","status-publish","format-standard","hentry","category-outils"],"acf":[],"_links":{"self":[{"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/posts\/41","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/comments?post=41"}],"version-history":[{"count":0,"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/posts\/41\/revisions"}],"wp:attachment":[{"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/media?parent=41"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/categories?post=41"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/numriq.com\/en\/wp-json\/wp\/v2\/tags?post=41"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}