v0.17.8

Ollama Releases / 3/13/2026

📰 NewsDeveloper Stack & InfrastructureTools & Practical UsageModels & Research

共有:

Key Points

The v0.17.8 release includes fixes across parsers, cloud model handling, docs formatting, localhost handling, and Docker build parallelism to improve stability and performance.
Parsers: repair unclosed arg_value tags in GLM tool calls, credited to BruceMacD in PR #14656.
Cloud models: reapply policy of not requiring pulling stubs for cloud models, as implemented by jmorganca in PR #14608.
Documentation: update and format compatibility docs to improve readability, contributed by mxyng in PR #14678.
Localhost and build: fix localhost handling in the create workflow and introduce smarter Docker parallelism to speed up builds, contributed by dhiltgen (PRs #14681 and related).

parsers: repair unclosed arg_value tags in GLM tool calls by @BruceMacD in #14656
Reapply "don't require pulling stubs for cloud models" again by @jmorganca in #14608
docs: format compat docs by @mxyng in #14678
create: fix localhost handling by @dhiltgen in #14681
build: smarter docker parallelism by @dhiltgen in #14653
mlx: int4 groupsize 64 by @pdevine in #14682
cloud_proxy: handle stream disconnects gracefully by @drifkin in #14685
x/mlxrunner: replace sampler interface chain with single stateful Sampler by @pdevine in #14652
rocm: update linux to v7.2 by @dhiltgen in #14391
app: fix reset to defaults disabling auto-update by @hoyyeva in #14741
mlx: get parameters from modelfile during model creation by @pdevine in #14747
MLX: add header vendoring and remove go build tag by @dhiltgen in #14642
ci: Fix windows build by @dhiltgen in #14754