Tag
Explore by tags
Runs a consumer AI chat interface backed by DeepSeek's large language models, with text, code, table, file, app, and API workflows. Its main appeal is strong reasoning access at unusually low cost.
Re-derives LLM scaling laws, tracing prior disagreements to how compute budget was modeled, then trains 7B and 67B models on 2T tokens. The 67B model beats LLaMA-2 70B on code, math, and reasoning; its chat variant tops GPT-3.5 on open-ended evals.
Reworks the MoE layer to push each expert toward a narrow specialty: split experts into many finer ones and activate more per token, plus reserve a few always-on shared experts for common knowledge. A 2B model matches GShard 2.9B; at 16B it rivals LLaMA2 7B on ~40% of the compute.
A family of open code models (1.3B-33B) trained from scratch on 2T tokens of project-level code, using a 16K-window fill-in-the-blank objective. Beats Codex and GPT-3.5 on code benchmarks and ships under a license permitting commercial use.
Reaches 51.7% on the competition-level MATH benchmark with a 7B model and no tools or voting, rivaling Gemini-Ultra and GPT-4. Built on a 120B-token math corpus mined from Common Crawl, and introduces GRPO, a memory-efficient PPO variant for reasoning.
Runs open LLMs entirely on your own machine — discover and download models from Hugging Face, chat in a desktop GUI, or expose an OpenAI-compatible local server. Native Apple MLX and llama.cpp backends; headless deploy via llmster.
A 236B-parameter Mixture-of-Experts LLM that activates only 21B parameters per token, cutting training cost 42.5% versus a dense 67B model and shrinking the KV cache 93.3% via Multi-head Latent Attention, with 128K context.
Offers OpenAI- and Anthropic-compatible access to DeepSeek models, including chat, reasoning, tool calls, JSON output, long-context variants, pricing, rate limits, and agent-tool integration guides.
Runs iterative, fully-local web research loops using locally hosted LLMs (via Ollama or LMStudio): it auto-generates search queries, gathers and summarizes results, reflects to find gaps, re-queries, and emits a final markdown report with sources.
A 671B-parameter Mixture-of-Experts language model (37B activated) trained on 14.8T tokens with 128K context, FP8-first training, a Multi-Token Prediction module, and Hugging Face weights—focused on efficient MoE training and long-context use cases.
Open-weight Mixture-of-Experts LLM with 671B total parameters but 37B activated per token, trained on 14.8T tokens for 2.788M H800 GPU-hours. Matches leading closed models at a fraction of typical training cost via FP8 and architectural tricks.
Shows that LLM reasoning can be incentivized through pure reinforcement learning, with no human-annotated reasoning traces. Self-reflection, verification, and strategy-switching emerge on their own, and the patterns transfer to distill smaller models.