OpenAI released GPT-6.1 Sol on September 29, achieving near-Astra results on agentic coding, computer use, and professional work tasks at one-fifth of GPT-6 Astra’s token price, according to (MarkTechPost). The model is priced at $2 input and $10 output per million tokens, with cached input dropping to $0.10, making it available for immediate use. In parallel, Perplexity Research and turbopuffer released pplx-embed-v2-context-9b-preview, a contextual embedding model designed for RAG pipelines that embeds each chunk with the full document in view, targeting improved answer retrieval with supporting evidence (MarkTechPost).
Google released Gemini 4 Argon, positioning it as a frontier model for coding and cybersecurity work with 1M output tokens (MarkTechPost). Independent testing shows Argon matches GPT-6 Astra on most benchmarks but consumes more than twice as many tokens per task, and access remains gated to select testers (The Decoder). Separately, Nvidia released Kumo Tabular, a family of tabular foundation models for classification and regression that predict new rows in a single forward pass without requiring training or hyperparameter tuning (MarkTechPost).