Open LLM Explorer · AI Agents · Benchmarks
Explore open models, calculate VRAM, & discover agents.
Comprehensive benchmarks, real-time GPU memory footprint calculator, context window boundaries, and curated open-source AI agent frameworks from the community.
22
Indexed Open Models
8
Curated AI Agents
160K
Max Context Length
100%
Open Source
Hardware Sizing Tool
LLM GPU Memory (VRAM) Calculator
Calculate precise GPU memory requirements for model weights, KV cache, and runtime activation buffers.
Estimated VRAM Required
5.8 GB
Includes weights + KV cache + 12% activation buffer
Model Weights
3.9 GB
KV Cache & Buffer
1.9 GB
Hardware Compatibility
💻 Entry / Ultrabook (4 GB RAM / iGPU)
✗ Out of Memory
💻 Desktop / Mac Base (8 GB - RTX 3050 / M2)
✓ Fits Easily
🖥️ Mid-Range GPU (16 GB - RTX 4070 / M-Pro)
✓ Fits Easily
🚀 Enthusiast Tier (24 GB - RTX 3090/4090 / M-Max)
✓ Fits Easily
🏢 Datacenter Cluster (80 GB - A100 / H100)
✓ Fits Easily
Open-Source Language Models
Open Model Leaderboard
| Model Name | Maintainer | Params | MMLU Score | Est. VRAM | Quantized | Context | License | Downloads |
|---|---|---|---|---|---|---|---|---|
| Qwen 32B | Qwen | 32B | 0.76 | 65.8 GB | ✓ | 128K | apache-2.0 | 679460 |
| Qwen2.5 Coder 32B Instruct | Qwen | 32B | 0.68 | 65.8 GB | ✓ | 32K | apache-2.0 | 389856 |
| DeepSeek V3 | deepseek-ai | 685B | 0.67 | 171.8 GB | - | 160K | Open | 775641 |
| Phi 4 | microsoft | 1.5B | 0.67 | 29.4 GB | - | 16K | mit | 422302 |
| Qwen 32B Preview | Qwen | 32B | 0.65 | 65.5 GB | ✓ | 32K | apache-2.0 | 201866 |
| DeepSeek R1 | deepseek-ai | 685B | 0.61 | 180.4 GB | - | 160K | mit | 1757133 |
| Llama 3.1 Nemotron Super 49B V1 | nvidia | 49B | 0.6 | 100.2 GB | - | 128K | other | 122265 |
| Mistral Small 24B Instruct 2501 | mistralai | 24B | 0.6 | 47.3 GB | - | proprietary | 918507 | |
| DeepSeek V3.0324 | deepseek-ai | 685B | 0.59 | 184.7 GB | - | 160K | mit | 248575 |
| DeepSeek R1 Distill Qwen 32B | deepseek-ai | 32B | 0.57 | 65.7 GB | - | 128K | mit | 2456452 |
| Bitnet B1.58 2B 4T | microsoft | 2B | 0.56 | 1.2 GB | - | 4K | mit | 17366 |
| DeepSeek R1 Distill Qwen 14B | deepseek-ai | 14B | 0.56 | 29.6 GB | - | 128K | mit | 914305 |
| Qwen2.5 Omni 7B | Qwen | 7B | 0.55 | 22.4 GB | - | 32K | other | 170998 |
| Kokoro 82M | hexgrad | 82M | 0.54 | 16 GB | - | apache-2.0 | 1981704 | |
| Phi 4 Mini Instruct | microsoft | 4B | 0.54 | 7.7 GB | - | 128K | mit | 312620 |
| Llama 3.1 Tulu 3 70B | allenai | 70B | 0.54 | 141.9 GB | - | 128K | llama3.1 | 13752 |
| Qwen2.5 7B Instruct 1M | Qwen | 7B | 0.54 | 15.4 GB | - | 986K | apache-2.0 | 2571997 |
| Mistral 8B Instruct 2410 | mistralai | 8B | 0.54 | 16.1 GB | - | MNPL-0.1 | 104428 | |
| Granite 8B Instruct | ibm-granite | 8B | 0.53 | 16.4 GB | - | 128K | apache-2.0 | 50356 |
| DeepCoder 14B Preview | agentica-org | 14B | 0.53 | 59.1 GB | - | 128K | mit | 35412 |
| DeepSeek R1 Distill Llama 70B | deepseek-ai | 70B | 0.53 | 141 GB | - | 128K | mit | 259503 |
| DeepSeek R1 Distill Qwen 1.5B | deepseek-ai | 2B | 0.53 | 3.5 GB | - | 128K | mit | 1699383 |
Autonomous Intelligence
Curated AI Agents & Frameworks
Open-source coding assistants, web browsing operators, multi-agent frameworks, and MCP orchestration runtimes.
| Agent | Type | GitHub Stars | Forks | License | Capabilities & Description |
|---|---|---|---|---|---|
| OpenClaw by OpenClaw Community | Ready-to-Use | 334,201 | 65,178 | MIT | Autonomous agent runtime with native tool execution and sandbox isolation. |
| Hermes Agent by Nous Research | Ready-to-Use | 219,347 | 41,593 | MIT | Advanced tool-use and autonomous function-calling agent powered by Hermes 3. |
| AutoGPT by Significant Gravitas | Ready-to-Use | 173,000 | 45,000 | Other | Pioneering autonomous agent architecture chaining LLM thoughts, commands, and memory. |
| Dify by LangGenius | Platform | 134,280 | 20,917 | Apache-2.0 | Open-source LLM application development platform orchestrating multi-agent workflows and RAG. |
| Gemini CLI by Google | Ready-to-Use | 106,138 | 14,301 | Apache-2.0 | Command-line agent for intelligent code refactoring, explainability, and terminal automation. |
| browser-use by browser-use | Framework | 84,000 | 9,700 | MIT | Make websites accessible for AI agents with automated DOM navigation and data extraction. |
| Flowise by Flowise | Platform | 51,044 | 23,976 | Apache-2.0 | Drag & drop visual builder for customized LLM flows, multi-agent systems, and vector search. |
| Goose AI by Block | Ready-to-Use | 14,200 | 1,150 | Apache-2.0 | Open-source, extensible developer agent that automates complex software engineering workflows. |