DeepSeek V4.1 Flash is a 552B-parameter mixture-of-experts model that activates only 8B-16B parameters per token and supports a 1M-token context window for handling long documents and agentic workloads. The model cuts KV-cache memory usage by 4x compared to its predecessor while improving coding, reasoning, and agent performance.

• Activates just 8B parameters during input processing and 16B during generation, making it efficient despite the massive 552B backbone
• Reduces memory footprint by roughly 4x versus DeepSeek-V4-Flash-0731 through Compressed Sparse Attention 2 and a Causal Encoder-Decoder architecture
• Offers configurable reasoning-effort levels so developers can adjust latency and cost versus reasoning depth
• Handles multimodal input (text and images) natively with a 1M-token context window, designed for agents processing large amounts of data

Laguna XS 2.1 is a 33B Mixture-of-Experts model with 3B activated parameters per token, tuned for agentic coding and long-horizon tasks on local machines. It outperforms XS.2 on SWE-bench Multilingual by 5.4 points, adds support for major runtimes, offers FP8/INT4/NVFP4 checkpoints, and runs with a permissive OpenMDW-1.1 license. You can access it via API, OpenRouter, or locally with quantized or DFlash-boosted builds.

• Laguna XS 2.1 jumps from 57.7% to 63.1% on SWE-bench Multilingual, beating Qwen 3.6 (35B) and North Mini Code (30B) despite only 3B active params per token
• DFlash speculator models double token throughput versus stock XS 2.1 when run locally
• Same API pricing as XS.2 ($0.10 input/$0.20 output/$0.05 cache-read per million tokens), now served at 256K context
• XS.2 leaves the API in one week, pushing users to migrate to 2.1 (though it remains available on Baseten's Model Library)

MIT CSAIL researchers built Retrieval Language Models that store full documents outside the model’s context window and let the AI query them via code, slicing, and parallel sub-instances. This approach handles inputs up to 10 million tokens, doubles benchmark performance, and matches or beats the cost of massive-context calls.

• MIT CSAIL's RLM approach stores full documents outside the context window and lets the AI query them via code (grep/slice) plus parallel sub-instances, instead of stuffing everything into context.
• It scales to inputs up to 10 million tokens, while standard LLMs like GPT-5 solve fewer than 10% of problems beyond 75,000 tokens.
• RLMs beat RAG methods by double-digit percentage gains at equal or lower cost per query.
• The code is open source and designed as a drop-in replacement for existing API calls, suggesting bigger context windows are a dead-end strategy compared to teaching models where to look.