Releases·Agent Release
Deep Dog 2: MIT-licensed Python deep research agent ranks #5 on DeepResearch Bench, defaulting to DeepSeek V4 Flash with Exa
A developer released Deep Dog 2, an open-source deep research agent claiming the best open-source performance and #5 overall on DeepResearch Bench. It installs via pip, runs asynchronously, and defaults to DeepSeek V4 Flash with Exa for search, returning reports as Markdown. Released free under the MIT License with no hosted service, so users pay only provider API costs.
4 engagement1 sourceSat, Sep 12, 2026, 02:00 PM
AI Overview
- Installs via pip from GitHub and exposes an async run_research() call returning a Markdown report; MIT-licensed with no hosted service
- Specialist agents include Web, PubMed, Reddit, Substack, and SEC Edgar; quickstart configures models, search engine, budgets, and iteration limits
- Default setup uses DeepSeek V4 Flash for supervisor, research sub-agents, and drafting, with Exa for web search; Tavily also supported
Related
Products·launched
Deep Dog 2: open source deep research agent that runs with multiple LLM providers
Deep Dog 2 is an open source deep research agent installable via pip that automates research tasks and reports outperforming Gemini and OpenAI on the DeepResearch bench at lower cost. It uses a variety of LLM providers (e.g. DeepSeek) plus Exa for search, configured via API keys. It targets developers who want a self-hosted research agent instead of closed paid services.
Deep Dog 2 is an open source deep research agent installable via pip that automates research tasks and reports outperforming Gemini and OpenAI on the DeepResearch bench at lower cost. It uses a variety of LLM providers (e.g. DeepSeek) plus Exa for search, configured via API keys. It targets developers who want a self-hosted research agent instead of closed paid services.
3 engagement1 sourceredditMon, Sep 14, 2026, 05:30 AM
Social·Community
6 DeepResearch API Compared: OpenAI, Perplexity, Parallel, Tavily, Gemini, Valyu Features and Pricing
A practitioner tested deep research APIs from OpenAI, Perplexity, Parallel, Valyu, Tavily, and Gemini, comparing tool control, deliverable generation, MCP support, review checkpoints, and per-task pricing. The writeup gives integration-oriented guidance for teams building research workflows, highlighting that pricing and capabilities diverge sharply across vendors.
A practitioner tested deep research APIs from OpenAI, Perplexity, Parallel, Valyu, Tavily, and Gemini, comparing tool control, deliverable generation, MCP support, review checkpoints, and per-task pricing. The writeup gives integration-oriented guidance for teams building research workflows, highlighting that pricing and capabilities diverge sharply across vendors.
5 engagement1 sourceredditFri, Oct 16, 2026, 02:04 PM
Products·launched
DeepSeek-powered deep research alternative with cached output
A side project that replicates deep research using DeepSeek, sending queries to multiple focused research agents then adversarial agents that challenge findings before synthesis. Built because commercial deep-research products (OpenAI, Claude) were too expensive and the author couldn't use a company LLM for personal use. It caches research output to avoid re-running expensive pipelines.
A side project that replicates deep research using DeepSeek, sending queries to multiple focused research agents then adversarial agents that challenge findings before synthesis. Built because commercial deep-research products (OpenAI, Claude) were too expensive and the author couldn't use a company LLM for personal use. It caches research output to avoid re-running expensive pipelines.
4 engagement1 sourceredditThu, Sep 3, 2026, 05:20 PM
Products·launched
Mole: Terminal deep research agent with budget and privacy controls
Mole is a terminal-based deep research agent that enforces a strict budget, verifies quotes with sources, and keeps local data private. It uses LLMs to conduct research tasks, ensuring cost control and data security for users who need reliable, private research.
Mole is a terminal-based deep research agent that enforces a strict budget, verifies quotes with sources, and keeps local data private. It uses LLMs to conduct research tasks, ensuring cost control and data security for users who need reliable, private research.
2 engagement1 sourcehackernewsFri, Aug 14, 2026, 06:52 PM
Products·building
Multi-Agent Research Assistant with Local LLMs via LangGraph
An enterprise multi-agent research assistant that uses LangGraph to orchestrate local LLMs (Qwen3, DeepSeek, Mistral) via Ollama. It performs web searches, analyzes information, and generates structured research reports with citations and quality metrics, solving the problem of automated research synthesis for knowledge workers.
An enterprise multi-agent research assistant that uses LangGraph to orchestrate local LLMs (Qwen3, DeepSeek, Mistral) via Ollama. It performs web searches, analyzes information, and generates structured research reports with citations and quality metrics, solving the problem of automated research synthesis for knowledge workers.
2 engagement1 sourceredditTue, Jul 21, 2026, 04:00 PM