Open-source AI supply chain

User-facing applications and their dependencies

376

nodes

1,449

edges

8

node kinds

16

edge kinds

Top stack by consumer category

One highest-ranked supply chain per consumer category, with full alternatives available by category.

7 categories

Coding Agent

100 ranked options, 4 apps, 84 models

Pareto
OpenHandsQwen2.5-Coder-7B-Instruct

Developer-facing agents that can modify repositories, run commands, or automate coding workflows.

78.6

overall

65.2

open

74.2

bench

12

non-open

Apache License 2.0, Docker/Moby, Hugging Face Hub, LiteLLM, MIT plus enterprise-restricted directories

view other options

Code Completion

86 ranked options, 3 apps, 77 models

Pareto
ContinueQwen2.5-Coder-7B-Instruct

Developer-facing assistants for inline coding help, IDE workflows, and code suggestions.

75.8

overall

59.8

open

81.3

bench

11

non-open

Apache License 2.0, Apple M5 Max, Continue, Creative Commons Attribution 4.0, Dev Containers

view other options

Chat Interface

80 ranked options, 8 apps, 64 models

Pareto
LibreChatQwen2.5-7B-Instruct

User-facing chat applications for local, self-hosted, or multi-provider language model access.

79.9

overall

73.1

open

74.2

bench

7

non-open

Apache License 2.0, Docker/Moby, Hugging Face Hub, LibreChat, LiteLLM

view other options

Agent Workspace

38 ranked options, 5 apps, 30 models

Pareto
LibreChatQwen2.5-7B-Instruct

User-facing workspaces for configuring, coordinating, or operating AI agents.

79.9

overall

73.1

open

74.2

bench

7

non-open

Apache License 2.0, Docker/Moby, Hugging Face Hub, LibreChat, LiteLLM

view other options

AI-powered Search

42 ranked options, 3 apps, 39 models

Pareto
AnythingLLMQwen2.5-7B-Instruct

Applications for retrieval, document question answering, and knowledge-base search.

79.5

overall

66.3

open

74.2

bench

7

non-open

AnythingLLM, Apache License 2.0, Apple M5 Max, Docker/Moby, Hugging Face Hub

view other options

Image Generation Platform

3 ranked options, 3 apps, 2 models

Pareto
Stable Diffusion web UIStable Diffusion XL Base 1.0

Creator-facing applications for image generation, image editing, and diffusion workflows.

68.6

overall

51.8

open

82

bench

11

non-open

AMD Instinct MI300X, Apache License 2.0, AWS Trainium3, BSD 3-Clause License, Diffusers

view other options

Local LLM Desktop

15 ranked options, 2 apps, 14 models

JanQwen2.5-7B-Instruct

Desktop applications for running local language models on consumer hardware.

71.9

overall

63

open

74.2

bench

7

non-open

Apache License 2.0, Apple M5 Max, Consumer device, Hugging Face Hub, Jan

view other options

Supply-chain graph

OpenHands with Qwen2.5-Coder-7B-Instruct

APPLICATIONMODELSOFTWAREDATASETBENCHMARKINFRASTRUCTURELICENSEsandboxed_bysandboxed_bysupportsdepends_onevaluated_onlicensed_astrained_ontrained_withhosted_bylicensed_asderived_fromtrained_ontrained_withevaluated_onhosted_bylicensed_asuses_datalicensed_aslicensed_asdepends_onlicensed_asimplementslicensed_asuses_dataOpenHandsopen coreQwen2.5-7B-Instructopen weightsQwen2.5-Coder-7B-Instructopen weightsSource-neededbenchmark dataunknownQwen2.5 trainingcorpusunknownSWE-bench taskinstancessource availablePyTorchopen sourceTransformersopen sourceLiteLLMopen coreOpenAI-compatible APIproprietaryOpenHands Runtimeopen coreDocker/Mobyopen sourceHumanEvalopen sourceSWE-benchsource availableApache License 2.0open sourceMIT plusenterprise-restrictedopen coreOther or compositelicenseunknownHugging Face Hubopen core
AI supply-chain graph

Top alternatives by layer

11 component layers, 10 alternatives each

Foundation and chat models

  1. Qwen2.5-7B-Instruct11,890,117
  2. Llama 3.1 8B Instruct10,031,112
  3. Mistral-7B-Instruct-v0.33,392,572
  4. Mixtral-8x7B-Instruct-v0.1824,791
  5. Phi-3-mini-4k-instruct610,575
  6. DeepSeek-R1-Distill-Qwen-7B489,121
  7. Gemma 2 9B IT394,810
  8. GPT-J-6B250,402
  9. Falcon-7B-Instruct243,678
  10. OLMo-2-1124-7B-Instruct52,510

Coding models

  1. Qwen2.5-Coder-7B-Instruct2,085,546
  2. Phi-3-mini-4k-instruct610,575
  3. CodeGen 350M Mono137,804
  4. StarCoder2-3B123,303
  5. DeepSeek Coder 6.7B Instruct120,203
  6. CodeLlama 7B Instruct28,105
  7. Granite 8B Code Instruct7,939
  8. CodeGemma 7B IT6,784
  9. SantaCoder4,201
  10. Magicoder-S-DS-6.7B303

Image and media generation models

  1. Stable Diffusion v1.51,704,765
  2. Stable Diffusion XL Base 1.01,436,868
  3. FLUX.1-dev984,011
  4. FLUX.1-schnell337,454
  5. Playground v2.5271,756
  6. DreamShaper 857,985
  7. Stable Diffusion 3.5 Large22,257
  8. PixArt-XL-2-1024-MS4,727
  9. Kandinsky 2.2 Decoder2,835
  10. HunyuanDiT v1.2369

Embedding and reranking models

  1. all-MiniLM-L6-v2230,826,064
  2. Nomic Embed Text v1.517,630,308
  3. BGE Large EN v1.514,158,114
  4. BGE Base EN v1.58,571,107
  5. Multilingual E5 Large7,017,018
  6. mxbai-embed-large-v14,752,072
  7. Jina Embeddings v33,461,822
  8. E5 Large v23,351,617
  9. GTE Large EN v1.51,550,013
  10. BGE Reranker Large1,534,971

Speech and audio models

  1. XTTS-v29,034,852
  2. Whisper large-v35,054,098
  3. Whisper small2,315,618
  4. wav2vec2-base-960h1,098,753
  5. Distil-Whisper large-v3933,990
  6. SeamlessM4T v2 Large410,860
  7. Parakeet TDT 0.6B v2360,108
  8. MusicGen Small186,078
  9. SpeechT5 TTS105,237
  10. Bark Small50,282

Inference and serving runtimes

  1. Ollama173,713
  2. llama.cpp115,795
  3. vLLM82,361
  4. LiteLLM49,818
  5. LocalAI46,758
  6. Ray42,824
  7. SGLang28,903
  8. Text Generation Inference10,860
  9. Triton Inference Server10,742
  10. KServe5,561

Training and fine-tuning frameworks

  1. Transformers161,459
  2. PyTorch100,623
  3. Unsloth66,142
  4. DeepSpeed42,488
  5. Diffusers33,817
  6. PyTorch Lightning31,178
  7. PEFT21,257
  8. TRL18,591
  9. Megatron-LM16,643
  10. Axolotl12,030

Hardware and accelerator infrastructure

  1. NVIDIA B200100
  2. NVIDIA H20099
  3. NVIDIA H10098
  4. NVIDIA GeForce RTX 509097
  5. AMD Instinct MI325X96
  6. AMD Instinct MI300X95
  7. Intel Gaudi 394
  8. Google Cloud TPU v6e Trillium93
  9. AWS Trainium392
  10. Apple M5 Max91

Agent sandboxing and execution isolation

  1. Kubernetes122,896
  2. OpenHands Runtime76,331
  3. Daytona72,445
  4. Docker/Moby71,643
  5. Firecracker34,843
  6. gVisor18,496
  7. Dagger15,920
  8. Kata Containers8,041
  9. Dev Containers5,500
  10. E2B Infra1,157

Evaluation tools and benchmarks

  1. OpenAI Evals18,645
  2. DeepEval16,056
  3. Ragas14,313
  4. LM Evaluation Harness12,897
  5. Prompt flow11,144
  6. OpenCompass7,075
  7. SWE-bench5,120
  8. MTEB3,295
  9. HELM2,819
  10. Lighteval2,439

Openness classes

open sourceopen coreopen weightssource availableproprietaryunknown

Node coverage

application15
model167
dataset11
software52
benchmark8
organization93
license16
infrastructure14

Highest adoption metrics in seed data

ComponentMetricValueSource-backed as of
all-MiniLM-L6-v2Hugging Face downloads230,826,0642026-06-10
Nomic Embed Text v1.5Hugging Face downloads17,630,3082026-06-09
BGE Large EN v1.5Hugging Face downloads14,158,1142026-06-10
Qwen2.5-7B-InstructHugging Face downloads11,890,1172026-06-09
Llama 3.1 8B InstructHugging Face downloads10,031,1122026-06-10
XTTS-v2Hugging Face downloads9,034,8522026-06-10
BGE Base EN v1.5Hugging Face downloads8,571,1072026-06-10
Multilingual E5 LargeHugging Face downloads7,017,0182026-06-10