Who contributed to Local LLMs on July 30, 2026?

16 developers shipped this update, including Daniel Hiltgen, Jesse Gross, Xuan-Son Nguyen, Titaniumtown, tnhnyzc, github-actions[bot], Fadi Arafeh, and eicherseiji, and 8 more.

What were the notable Local LLMs updates?

CI: enable lint on the whole tree (#17457), lint: clean up current tree, and mlxrunner: count every speculative round and log stats at info.

Local LLMs

Ollama, llama.cpp, vLLM, and SGLang - running and serving open models locally

Pick a date

Topics: Local LLMs AI / ML Full archive →

The Wire · Showcase

OLLAMA CLEANS UP LINT PIPELINE AS LLAMA.CPP ADDS SPECULATIVE DRAFTING FOR GLM-5.2

By RepoJournal · Filed 07:02 UTC on July 30, 2026 · About Local LLMs

16 people shipped this

Daniel Hiltgen @dhiltgen 2 cited

github-actions[bot] @github-actions[bot] 2 cited

Jesse Gross @jessegross 1 cited

Titaniumtown @Titaniumtown 1 cited

tnhnyzc @tnhnyzc 1 cited

Xuan-Son Nguyen @ngxson

Fadi Arafeh @fadara01

eicherseiji @eicherseiji

BugenZhao @BugenZhao

JaredforReal @JaredforReal

Ollama's CI now scans the entire codebase instead of just changed lines, unblocking larger PRs that were silently accumulating lint violations across their diffs.

The shift fixes a critical gap in golangci-lint's only-new-issues mode, which filters findings down to touched lines and drops issues introduced at a distance where the report anchors to an untouched line [1]. Daniel Hiltgen's follow-up PR cleans the tree of existing violations so dependent work lands without carrying unrelated fixes [2]. Meanwhile, Jesse Gross improved speculative decoding diagnostics by logging per-request stats at info level instead of debug, capturing every speculative round to surface throughput issues [3]. Across llama.cpp, the project shipped two releases with expanded speculative decoding support: GLM-5.2 now has NextN/MTP drafting via the --spec-type draft-mtp target [4], and RDNA3.5 tuning configs can now be optimized independently [5]. On the performance front, a new SYCL contiguous fast path eliminates redundant 64-bit integer math from unary elementwise ops, cutting SILU kernel time from 3.2 ms down toward the 0.63 ms bandwidth floor [6]. The same release cycle adds MTP draft support for mimo2 architecture, preserving trunk-only GGUFs while building the MiMo-specific graph with fused attention and shared-head fallbacks [7].

Action items

→ Land dependent PRs against ollama/ollama now that CI lint cleanup is merged ollama/ollama [plan]
→ Test GLM-5.2 speculative decoding on llama.cpp if you serve Chinese reasoning models ggml-org/llama.cpp [monitor]
→ Pull SYCL fast-path release (b10175) for Intel Xe GPU deployments ggml-org/llama.cpp [plan]

References

[1] CI: enable lint on the whole tree (#17457) ollama/ollama
[2] lint: clean up current tree ↗ ollama/ollama
[3] mlxrunner: count every speculative round and log stats at info ollama/ollama
[4] b10174 ↗ ggml-org/llama.cpp
[5] b10175 ↗ ggml-org/llama.cpp
[6] sycl: contiguous fast path + 32-bit index math for unary elementwise ops ↗ ggml-org/llama.cpp
[7] mimo2: add MTP draft support ↗ ggml-org/llama.cpp

Quick answers

What shipped in Local LLMs on July 30, 2026?: Ollama's CI now scans the entire codebase instead of just changed lines, unblocking larger PRs that were silently accumulating lint violations across their diffs. In total, 113 commits, 113 pull requests, and 9 releases landed.
Who contributed to Local LLMs on July 30, 2026?: 16 developers shipped this update, including Daniel Hiltgen, Jesse Gross, Xuan-Son Nguyen, Titaniumtown, tnhnyzc, github-actions[bot], Fadi Arafeh, and eicherseiji, and 8 more.
What were the notable Local LLMs updates?: CI: enable lint on the whole tree (#17457), lint: clean up current tree, and mlxrunner: count every speculative round and log stats at info.

OLLAMA FIXES KV CACHE CORRUPTION, LLAMA.CPP HARDENS FILE SEARCH, VLLM KILLS MOE LEGACY CODE

Ollama patched a critical cache corruption bug that corrupts lazy snapshots during multi-request prefills, while llama.cpp hardened Windows junction traversal and vLLM removed 14 hardcoded MoE shards across quantization methods.

+10

llms 240 shipped

Local LLMs Aug 5

OLLAMA FIXES RACE CONDITIONS AS LLAMA.CPP SHIPS BREAKING TTS CHANGES

Ollama patched critical data races across its progress tracking system while llama.cpp introduces incompatible Qwen3-TTS support, and vLLM lands kernel optimizations that cut latency on specialized hardware.

+11

llms 246 shipped

Local LLMs Aug 4

OLLAMA REWIRES OPENAI COMPATIBILITY, LLAMA.CPP HARDENS SAMPLING

Ollama reworked its entire streaming protocol to match OpenAI's wire format chunk-for-chunk, unblocking drop-in replacement deployments.

+11

llms 230 shipped

Local LLMs Aug 3

DEEPSEEK V4 INFERENCE GETS 50% SPEEDUP; LLAMA.CPP AND VLLM BOTH SHIPPING OPTIMIZED PATHS

DeepSeek V4 speculative decode just went from nice-to-have to standard: llama.cpp landed MTP and DSpark support with roughly 50% speedup [ref:3], vLLM hardened its KV cache coordinator for the same models [ref:7], and Metal got native hyper-connection ops [ref:5].

llms 148 shipped

Elsewhere on the wire

AI Agents Aug 6

LANGCHAIN HARDENS ANTHROPIC TOOL CALLING AND GIT CREDENTIALS AS OPENHANDS SHIPS 1.10.0

LangChain fixed a critical credential leak in evaluation runs while Anthropic tool schemas get proper fallback handling, as three major agent frameworks push releases across the ecosystem.

ai-agents 54 shipped

@CachyOS Aug 6

WIKI TRANSLATION ROLLBACK CHAOS, MIRROR STACK HOLDING STEADY

CachyOS wiki maintainers are caught in a revert loop over German translations while the status infrastructure logs 55 commits without incident.

infra 57 shipped

Elixir & Phoenix Aug 6

ELIXIR BUMPS ZIZMOR SECURITY SCANNER TO 1.28.0

The Elixir repo updated its CI security action overnight, pulling in the latest zizmor static analysis tooling.

elixir 1 shipped

@dotnet Aug 6

DOTNET FOUNDATION PULLS iOS METRICS INTO FOCUS AS RUNTIME HARDENS HTTP PARSING

Performance team fixed iOS size reporting to measure only distributable packages, while runtime clamped down on RFC 9112 compliance for chunked encoding.

dotnet 72 shipped

Want every project, not just this one?

Follow Local LLMs