Pin llama.cpp v1.0.0: Major Version Bumps Now Mean ABI Breaks
llama.cpp and ggml both reached v1.0.0 on 2026-11-13. Major bumps now signal C API/ABI breaks. Here are the semver rules, packaging channels and what to pin.
33 posts
llama.cpp and ggml both reached v1.0.0 on 2026-11-13. Major bumps now signal C API/ABI breaks. Here are the semver rules, packaging channels and what to pin.
CVE-2026-119618 is an unauthenticated RCE in Apache HTTP Server's mod_http2, affecting 2.4.55 through 2.4.70. Upgrade to 2.4.71 or set Protocols http/1.1.
OpenSSH 10.6p2 fixes SSHhole: on musl systems like Alpine, any 257-1024 byte password logs into key-only accounts. Fixed package versions and a stopgap config.
encoding/json/v2 and jsontext are stable in Go 1.27: Unmarshal is up to 10x faster, and defaults now reject duplicate names and invalid UTF-8. Migration options inside.
IBM's Granite Speech 5.0 2B scores 5.06% average English WER under Apache 2.0, TurboCTC hits 19,870 RTFx, and 4.1 2B Plus halves timestamp error.
Recursive's Flint is a 39.8B MoE with ~2.5B active params, 262K context, Apache 2.0 weights and free OpenRouter access. Scores are self-reported.
DeepSeek open-sourced DeepInfer, its production inference framework, under Apache-2.0. It is offline-only, SM100/Ascend 950 only. Env vars and gaps inside.
DeepSeek-V4.1-Flash-1017 (MIT, 763B params) lifts Terminal-Bench 4.0 from 31.2 to 42.3. Benchmark table, serving commands, sampling settings and missing details.
Meta's 816B-param, 42B-active multimodal MoE runs on two M5 Ultra Mac Studios with 1M context, Helion kernel speedups, and open weights via MLX, vLLM, and SGLang.
OpenCode 2.0 ships a JS Code Mode, background subagents, a default-on server process and an SDK. Breaking changes: 1.0 plugins and the old API are gone. Here's what actually changed.
Liquid AI's LFM2.5-24B-A2B packs a 24B MoE with only 2B active parameters, 131K context, and 184 tok/s on an M5 Max — enough to keep agent loops fully local.
Mistral Large 4 public preview: 1T-parameter multimodal MoE, 49B active, $1.36/$4.18 per million tokens, weights promised by end of October.
Reflection's Beam is a 501B MoE with 23B active parameters, 80.1 on Terminal Bench 2.1 and 80.9 on SWE-bench Verified. Apache 2.0 weights this month.
Strata runs Qwen3.8-Flash-Next, a 125B MoE, on 12GB consumer cards: 94 tok/s write speed at Q2_0 on an RTX 5070, with an OpenAI/Anthropic-compatible server on localhost.
Hugging Face's tokenizers v1 release candidate encodes 3-30× faster than v0.23 with identical token IDs, via bitstream splitting, word caching, and a no-alloc merge loop.
NVIDIA's CUDA Rust gives Rust developers two tracks — SIMT via cuda-oxide and Tile via cutile-rs — to write GPU kernels natively in Rust, compiled to PTX, with compile-time memory safety.
Asahi Linux officially supports M3 Macs via the installer's Expert mode. Everything works except the GPU and sleep — here's what to expect.
Quantization-Aware Healing distills a 4-bit model from the pre-compression teacher, winning 7 of 9 benchmarks over its bf16 source.
IBM's Granite 4.2 — 3B/8B/30B, Apache 2.0, 512K context, thinking/non-thinking switch, and agentic RL on 8B/30B with real SWE, terminal, and search environments.
Qwen 3.8-27B is the new dense flagship you can run on a single 24-32 GB GPU. It jumps to 73.0 on Terminal Bench 2.1 (up from 63.4), adds native image and hour-scale video understanding, and ships under Apache 2.0. What changed versus Qwen3.6-27B, how to run it at 4-bit, and the catches I would watch.
vLLM v0.27.0 lets hybrid SWA+full attention models run through offline inference, closing the gap between eval and serving — plus a breaking torch 2.13 upgrade.
Microsoft's Agent Framework Harness and Foundry Hosted Agents hit GA August 3 — a consumption-billed runtime with per-session isolation and scale-to-zero. Here's what it costs and where CodeAct fits.
Microsoft Agent Framework's new Harness collapses planning, memory, compaction, and approvals into one call in Python and .NET.
vLLM v0.26.0 adds runtime draft-weight updates, so you can retune speculative decoding without restarting the engine and flushing the KV cache.
Microsoft Agent Framework now backs agents with GitHub Copilot SDK, shipping shell execution, file ops, and MCP integration in stable C# and Python SDKs.
CVE-2026-59726 lets unauthenticated attackers execute shell commands, steal LLM keys, and poison agent memory via Ruflo's exposed /mcp endpoint. Upgrade to 3.16.3 and audit immediately.
NVIDIA embeds sparse linear algebra solvers into the Agent Toolkit—cuISS, cuDSS, cuEST—so autonomous design agents call physics simulation without leaving the flow. Free libraries, GPU-only execution.
MCP protocol drops session-based architecture July 28, making agents scale horizontally without sticky routing. New SDKs support both old and new versions, but sampling deprecation requires network auth changes.
TypeScript 7.0 stable (July 8) ships a native Go compiler that slashes build times by 8–12× — but Vue, Svelte, and Astro lose template type-checking until the programmatic API stabilizes in 7.1.
Tencent's Hy3 benchmarks between models 2–5x larger by parameter count. 256K context, Apache 2.0, 1 yuan per million input tokens. The catch: vendor benchmarks need independent testing.
University of Waterloo's PAW system compiles fuzzy-function specs into 23MB LoRA adapters that match 32B model accuracy on a 600M interpreter—no API calls, no cloud dependency, 30 tok/s on MacBook M3.
A2UI v0.9 flips the script: agents declare UI intent against your existing design catalog, not invent components. What changes for agentic architectures.
A VS Code extension that brings Azure DevOps sprint boards, work item management, and AI-powered assistance directly into your editor.