Show HN: ThinkLLM, A knowledge graph of AI models (HTTPS://thinkllm.dev)
ThinkLLM is a living knowledge graph that helps you browse, compare, and find AI models by task and capabilities.
Article intelligence
Key points
- ThinkLLM organizes thousands of AI models by task and capability, simplifying model selection.
- Categories include agentic tasks, coding, content creation, customer support, data analysis, and more.
- Models are tagged with metadata such as open weight status and recency.
Why it matters
This matters because thinkLLM organizes thousands of AI models by task and capability, simplifying model selection.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
ThinkLLM — Think through your LLM choices
Think through your LLM choices
A living knowledge graph of AI models. Browse by task, compare capabilities, and find the right model for your project.
Browse Use CasesExplore ModelsCompare Models
What are you building?
View all →
Start here to find models that fit your task
Capabilities
View all →
Recently Added
View all models →
Agentic Tasks
207 models
Multi-step tasks requiring planning, tool use, and self-correction
Research a topic and compile a structured reportBrowse the web to complete a booking task
Coding Assistant
466 models
Writing, reviewing, debugging, and explaining code in a development workflow
Generate a REST endpoint from a descriptionReview a PR and suggest improvements
Content Creation
589 models
Generating marketing copy, articles, social posts, and creative content
Write a product description for an e-commerce listingDraft a blog post outline and first draft
Customer Support
518 models
Handling customer inquiries following product policies and tone guidelines
Answer a billing question using a policy documentEscalate a complaint with the correct tone
Data Analysis
421 models
Interpreting data, writing analysis code, and explaining findings
Write a pandas script to clean and aggregate a CSVExplain what a regression output means
Local Deployment
1365 models
Running open-weight models on local hardware — privacy, cost, or offline use
Self-hosted chatbot on consumer GPUOffline inference for sensitive data
RAG & Retrieval
723 models
Answering questions grounded in a retrieved document corpus
Answer questions from internal documentationSummarize relevant sections of a PDF
Research Synthesis
448 models
Reading, summarizing, and connecting information across multiple sources
Summarize the key findings of 5 papers on a topicCompare two competing theories
Coding
836 models
Code generation, debugging, explanation, and refactoring
Generate working functions from specsDebug a failing test
Creative Writing
917 models
Creative, narrative, and stylistic text generation
Write a short story with a specific tone and structureGenerate marketing copy in a brand voice
Factual Knowledge
1111 models
World knowledge accuracy, recall of facts and relationships
Answer factual questions about history, science, geographyRecall accurate dates, names, and relationships
Instruction Following
1099 models
Adhering to complex, structured, or constrained instructions
Follow a multi-step formatting template exactlyRespect negative constraints (do not mention X)
Long Context
1037 models
Performance retention over long documents and conversations
Answer questions about a 100-page documentMaintain consistent character over a long conversation
Multilingual
934 models
Quality of non-English language understanding and generation
Answer questions in French, Spanish, Chinese, etc.Translate with idiomatic accuracy
Multimodal
328 models
Quality of vision, audio, and image understanding (distinct from modality support)
Describe the contents of a complex imageInterpret charts and diagrams accurately
Reasoning & Logic
917 models
Multi-step reasoning, logic puzzles, mathematical problem-solving
Multi-step math word problemsDeductive and inductive logic puzzles
Tool Use
512 models
Function calling, structured output, agent-style tool orchestration
Call the right API tool with correct parametersChain multiple tool calls to complete a task
HRM Text 1BOpen WeightModel weights are publicly available — can be downloaded and self-hosted
3d ago
Gemma4 26BA4B · Uncensored · HauhauCS · BalancedOpen WeightModel weights are publicly available — can be downloaded and self-hosted
9d ago
granite embedding 311m multilingual r2Open WeightModel weights are publicly available — can be downloaded and self-hosted
IBM
10d ago
Qwen3.5 9BFP8 · dynamicOpen WeightModel weights are publicly available — can be downloaded and self-hosted
12d ago
Kimi K2.6NVFP4Open WeightModel weights are publicly available — can be downloaded and self-hosted
NVIDIA
12d ago
jina embeddings v5 omni nanoOpen WeightModel weights are publicly available — can be downloaded and self-hosted
15d ago
Qwen3.6 35BA3B · uncensored · heretic · Native · MTP · PreservedOpen WeightModel weights are publicly available — can be downloaded and self-hosted
15d ago
jina embeddings v5 omni smallOpen WeightModel weights are publicly available — can be downloaded and self-hosted
15d ago