DeepSeek V4.1 Flash 1017 is a snapshot release (Oct 17, 2026) of a sparse mixture-of-experts model built on DeepSeek’s Causal Encoder-Decoder architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, using an asymmetric activation split to keep per-token compute low. Visual and text embeddings were trained jointly from pretraining, making image understanding native to the model. It supports a 1,048,576-token context window (with up to 393,216 completion tokens), accepts text and images as input and returns text, and offers tool calling (tools, tool_choice) plus a JSON response_format option without schema enforcement. Published pricing is $0.15 per 1M input tokens, $0.60 per 1M output tokens, and $0.003 per 1M cache-read tokens.
Targeted at coding, terminal/computer-use agents and long-horizon, multi-step workflows, the model emphasizes cost-efficient agentic use: compressed KV caching reduces cache memory to roughly a quarter of the prior Flash generation. DeepSeek positions this tier as a cost-efficient V4.1 offering and claims improvements over V4 Pro in performance, speed, and task completion time. Measured throughput is about 155 tokens/sec with P50 latency ~1.24s and near-100% uptime/99.72% availability over three days. The model is hosted by a single provider via OpenRouter and currently lacks EU in-region routing.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
DeepSeek-v4.1-Flash-1017 is a multimodal image-text-to-text model available on Hugging Face, optimized for text generation tasks. It supports various precisions and is licensed under MIT, with deployment options in the US region. (huggingface.co)
Meta plans to scrape content from its business customers to train its AI models. This practice raises concerns about data privacy and user consent. (theregister.com)
Ray Myers criticizes OpenAI for prioritizing profit over genuine mathematical progress, accusing it of dishonesty and devaluing workers. He compares the AI industry’s behavior to narcissistic abuse, emphasizing a toxic relationship with artificial intelligence that benefits only the dominant players. (raymyers.org)
Llmspeed allows users to simulate the speed of local large language models on various hardware setups before making a purchase. It provides benchmark estimates based on community data for different models and devices. (llmspeed.dev)
Elon Musk has made numerous predictions about Tesla achieving full autonomous driving, with timelines ranging from 2018 to 2023. As of 2024, Tesla's FSD system remains classified as Level 2, and full autonomy has not yet been achieved. (en.wikipedia.org)
Lithos-metal develops high-performance LLM inference kernels and serving solutions optimized for Apple silicon. The project is hosted on GitHub with 257 stars and 24 forks, focusing on efficient AI model deployment. (github.com)
Today's best Hacker News stories, summarized and screenshotted, one email a day.