Browse the docs · Apsi 2.0.0

Versions: what each one can do

	2.0.0 (current)	1.10.0
Released	November 2026	8 November 2026
Engines	bytecode engine + native compiler (`@native`, LLVM machine code), `parallel for`, NVIDIA GPU kernels	bytecode engine
Speed on the benchmarks	next to C, C++, Rust; ahead of Go	about 500× slower than 2.0.0 (geometric mean; 12× to 2,400× per program)
Errors	`try / except / else / finally`, `raise`, Python kind names	`error("…")` stops the program
Generators, `args/*kwargs`, operator methods on structs	yes	no
Python conversion	keeps exceptions, generators, classes with operators, `for … else`, pure stdlib modules	core language and numpy; exceptions become comments
PyTorch	exact conversion of whole training programs, native layers/optimizers/schedulers	`use torch` layers, conversion of models
Reinforcement learning, science modules	`use rl`, special functions, statistics, DSP, PDE, quantum, mechanics, medical, lab	AI and physics basics
Large models	`apsi plan`, model files (GGUF, safetensors, `.apsim`), sharding	—
Libraries (`apsi install / pack / publish / yank / unpublish`)	yes	yes
Website commands (`apsi release …`)	yes	yes
Standard modules	34 modules, 833 members, 253 built-ins	22 modules, 360 members, 201 built-ins
Documentation	2.0.0 docs	1.10.0 docs

Which one should I use?

2.0.0 for everything new. Programs written for 1.10.0 run on 2.0.0 unchanged — 2.0.0 only adds.

Keep 1.10.0 only if a machine cannot install 2.0.0's native compiler dependency (llvmlite); even then 2.0.0 works without it, on the bytecode engine, so the reasons are rare.

Moving from 1.10.0 to 2.0.0

Install 2.0.0 (it replaces 1.10.0).
Run your tests: apsi test.
Add @native to the functions that do the heavy loops — see maximum performance.
If you used error() for things you would like to recover from, switch to raise and try / except.

Found a mistake? Report it · Markdown source