← Back to Research Catalog
USEReleased: 8/14/2026• Evidence: HIGH

Ollama v0.32.13 Released: Developer Instructions Support for Qwen 3.8

Ollama release v0.32.13 introduces support for developer instructions in Qwen 3.8 models alongside multi-platform binary distributions for Darwin, Linux, and Windows.

What Changed

Added developer instruction parameters for qwen3.8 model execution. Updated pre-built installation archives and binaries across Linux amd64/arm64, macOS Metal/Darwin, and Windows ROCm/MLX targets.

Why It Matters

Developer instructions provide system-level behavioral constraints for local agent reasoning loops without relying on user prompt context.

Anti-Signal / Limitations & Skepticism

Developer instruction parsing requires compatible Qwen 3.8 GGUF model variants; older Qwen 2.5 or un-patched fine-tunes ignore developer instruction parameters. Memory footprint for Qwen 3.8 32B remains ~22GB VRAM.

Primary Sources & Evidence

Technical Stack

OllamaQwen 3.8GGUFllama.cpp

Hardware Compatibility

Mac x86_64 32GBApple Silicon M-SeriesLinux CUDA