Free Developer Radar & Open Weights Directory
High-performing open-weights models you can execute locally or access with generous free API tiers. No hidden subscription walls, no fake free trials, with real-time GitHub telemetry.
🤖 Top Open-Weights AI Models (Local Execution & Free APIs)
DeepSeek-R1 & DeepSeek-V3
★ 91,990Breakthrough open-weights reasoning model with verified chain-of-thought capabilities outperforming commercial closed LLMs on coding and mathematics.
Qwen 2.5 Coder 32B
★ 16,840Alibaba Cloud's flagship open-weights coding model supporting 92 languages, repository-scale reasoning, and AST repair.
Llama 3.3 70B Instruct (Meta)
⚡ Verified Free QuotaMeta's flagship open-weights model delivering GPT-4o-level performance with 128k context window and rich multilingual fluency.
⚡ Generous Zero-Cost Developer APIs
Groq Cloud Fast LPU Inference API
⚡ Verified Free QuotaUltra-low latency inference engine powered by custom Language Processing Units (LPUs). Token generation up to 1,200 tokens/sec for open models.
Google AI Studio API (Gemini Flash)
⚡ Verified Free QuotaDeveloper API access to Gemini 2.5 and Flash models with native 1M token context window, audio, vision, and multimodal grounding.
Cloudflare Workers AI
⚡ Verified Free QuotaServerless GPU inference running on Cloudflare's global edge network. Access Llama 3, Mistral, and BGE vector embeddings directly in edge workers.
📦 Essential Open Source Infrastructure Repos
Get up and running with large language models locally. Lightweight CLI, REST API, and native hardware acceleration across macOS, Linux, and Windows.
Thousands of pretrained models to perform tasks on texts such as classification, information extraction, question answering, and summarization.
State-of-the-art serving engine utilizing PagedAttention, continuous batching, and chunked prefill. 2-4x higher throughput than standard runtimes.