The state of the Python ecosystem in 2026 H2

Quick answer. Python is still the most-tracked language on the GitPop radar in 2026 H2, with 16,800+ active projects. The growth is concentrated in AI/ML: LangChain, LlamaIndex, vLLM, llama.cpp, Ollama, and the rest of the LLM-tooling ecosystem. Outside AI, the data tooling (Polars, DuckDB, Pandas 3) is the next-fastest-growing segment. The web framework segment is flat — Django, FastAPI, and Flask are stable but not growing.

The 2026 H2 snapshot

The GitPop radar tracks 16,800+ active Python projects in 2026 H2, the most of any language. The total project count is up 12% year over year, and the PopScore of the average tracked Python project is up 9%.

The growth is uneven. The AI/ML segment is up 64% year over year. The data tooling segment is up 22%. The web framework segment is up 1%. The CLI tooling segment is down 4% (Rust is winning CLI work; see The state of the Rust ecosystem in 2026 H2).

Where Python dominates

AI / ML infrastructure

The single biggest growth segment in 2026. The GitPop radar tracks the leading Python AI/ML projects:

• huggingface/transformers — model library (PopScore 95)
• langchain-ai/langchain — LLM orchestration (PopScore 89)
• vllm-project/vllm — LLM serving (PopScore 84)
• run-llama/llama_index — RAG framework (PopScore 78)
• ollama/ollama — local LLM runtime (PopScore 91)
• ggerganov/llama.cpp — inference engine (PopScore 87)
• BerriAI/litellm — LLM proxy (PopScore 69)
• crewAIInc/crewAI — multi-agent framework (PopScore 71)
• langchain-ai/langgraph — stateful agents (PopScore 82)

Python is the default language for AI/ML work. The pattern is consistent: a Python API, with a Rust core for performance-critical paths. vLLM (Python + CUDA), llama.cpp (C++ with Python bindings), and Ollama (Go with Python bindings) all follow this pattern.

For the broader AI toolchain, see Building your AI dev stack in 2026 and Best open source AI tools in 2026.

Data tooling

The next-fastest-growing Python segment. The leading projects:

• pandas-dev/pandas — DataFrame library (PopScore 88)
• pola-rs/polars — Rust-core DataFrame (PopScore 78)
• duckdb/duckdb — in-process OLAP database (PopScore 83)
• numpy/numpy — array library (PopScore 91)
• scikit-learn/scikit-learn — classical ML (PopScore 87)
• matplotlib/matplotlib — plotting (PopScore 82)

The interesting story here is the rise of Polars. The Rust-core, Python-bindings pattern is winning in data tooling. Pandas 3 (released in 2025) is faster but still has the legacy pandas API. For new projects in 2026, Polars is the default. See Pandas vs Polars for the head-to-head.

DuckDB is another standout. The in-process OLAP database has gone from "interesting experiment" to "default for analytical Python work" in 2025-2026.

Scientific computing

Python is still the default for scientific computing, but the segment is flat to declining as Rust-native alternatives (Polars, DuckDB, ndarray) capture share:

• numpy/numpy — array library (PopScore 91, flat)
• scipy/scipy — scientific computing (PopScore 84, flat)
• sympy/sympy — symbolic math (PopScore 71, flat)

These are mature, stable projects. The PopScore is high, the growth is low. This is the "infrastructure tier" of the Python ecosystem.

Where Python is stable

Web frameworks

Django, FastAPI, and Flask are the default Python web frameworks in 2026. The PopScores are high (Django 89, FastAPI 91, Flask 84) but the growth is flat.

The reason: the web framework segment in 2026 is dominated by JavaScript (Next.js, Astro, Remix) and TypeScript-first stacks. Python web frameworks are used for ML-backed APIs (FastAPI is the default for serving ML models) and for content-heavy sites (Django is the default for CMS-style apps). For new product work, JavaScript is still the default.

See Pick a Python web framework in 2026 for the framework decision tree.

CLI tooling

The CLI segment is the one where Python is losing share. The new generation of CLI tools (ripgrep, fd, bat, eza, zellij) is Rust-native. Python CLIs are still common (httpie, poetry, pip) but the high-momentum new tools are Rust.

The reason: Python's startup time and distribution model (need to install a Python interpreter, plus dependencies, before you can run the tool) are friction points. A single Rust binary that just works is a better user experience for CLI tools.

What is next

The Python ecosystem is in a stable-but-evolving phase. The 2026 H2 outlook:

1. AI/ML will continue to grow. The LLM ecosystem is not slowing down. The Python-native projects in this space will continue to gain PopScore.
2. Data tooling will continue to evolve. Polars, DuckDB, and the Rust-core-with-Python-bindings pattern will capture more of the data tooling market.
3. Web framework segment will stay flat. Python web is not growing. New product work is still JavaScript-first.
4. CLI will continue to lose share to Rust. The new generation of CLI tools is Rust-native. The trend will continue.
5. The Polars/DuckDB pattern will spread. Rust-core, Python-bindings is the default for high-performance Python tools going forward. Expect more libraries to follow this pattern in 2026-2027.

For the broader data, see State of GitPop H2 2026. For the comparison with the Rust ecosystem, see The state of the Rust ecosystem in 2026 H2. For the framework decision, see Pick a Python web framework in 2026.