AI INFRASTRUCTURE / MODEL RUNTIMES
Run models on your terms.
Compare local inference libraries, packaged runtimes, and their documented limits.
All spaces · 4 targets · 14 capabilities in 4 groups · 42 claims · 21 sources · last capture 2026-09-15 · 4 rechecks due
MODEL RUNTIMESSOFTWARE, IN CONTEXT.
4 of 4 targets
Documentation-only research. Hardware, model formats and bundled engines affect support; speed, memory fit and output quality have not been measured. Missing evidence remains Unknown.
0 fully sourced · 4 partly sourced · 0 not yet sourced
| Local inference | Apple Silicon acceleration | CPU inference | NVIDIA acceleration | GGUF models | MLX models | Quantized weights | Local HTTP server | OpenAI-style chat endpoint | Streaming generation | JSON schema output | Concurrent inference | Prompt cache reuse | Speculative decoding | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Fully sourced0 of 4Every capability in this space carries at least one cited claim for this target. | |||||||||||||||||
| Partly sourced4 of 4Some capabilities carry cited claims; the remainder have no cited claim in this inventory. | |||||||||||||||||
| 13 / 14 | 2 | due | |||||||||||||||
| 10 / 14 | 3 | due | |||||||||||||||
| 8 / 14 | 3 | due | |||||||||||||||
| 11 / 14 | 2 | due | |||||||||||||||
| Not yet sourced0 of 4No capability in this space has a cited claim in this inventory for this target. That is a missing citation, not evidence the capability is absent. | |||||||||||||||||