Edition of Saturday 14 November 2026

llama.cpp v1.0

Comments on “llama.cpp v1.0”

Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked

llama.cpp and ggml both reached v1.0.0, and the maintainers say API and ABI breaks will now be signaled with major-version bumps. The release removes 35 deprecated functions from llama.h, moves the NextN/speculative API into it, and adds FP8 KV caches, a host-memory MoE expert cache, new model architectures, a /v1/audio/speech endpoint and llama-server security fixes (SSRF, CORS, DNS rebinding). Downstream builders should check for removed calls before upgrading. The details come from the project's release notes; the claims were not independently verified.

Read the original at github.com. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Why it ranked

No. 4 of 123 with a score of 0.75. The panel rated it highest on new information (0.92) and technical depth (0.88), lowest on audience breadth (0.66).

AI judgeJudge 2claude-haiku-5-5
Major release of a widely used local LLM inference library with API breaks, FP8 KV caches, server security hardening; detailed changelog, primary source, but largely a release-notes list with little explanation.
AI judgeJudge 3gpt-5.6-luna
A detailed, primary release covering major llama.cpp and ggml API, inference, multimodal, backend, server-security, and UI changes. It is highly actionable and well evidenced, though mostly a changelog rather than explanatory engineering analysis.

Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.

Score

0.75total under rubric v2.3

Judges’ mean answer per question, on a scale from 0 to 1. Questions marked minus count against the score.
  • Significance0.71
  • Reader interest0.77
  • Audience breadth0.66
  • Practical relevance0.85
  • Substantiated0.86
  • Technical depth0.88
  • New information0.92
  • Counts against: Promotional0.06
  • Counts against: Overclaim0.14

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

QuestionJudge 1Judge 2Judge 3MeanSpreadWeightContribution (weight times mean)
SignificanceHow much does this change what people in tech know or can do?0.760.550.830.710.28+0.24+0.171
Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on?0.720.700.900.770.20+0.26+0.201
Audience breadthHow many people who build technology does it concern?0.540.720.720.660.18+0.13+0.086
Practical relevanceHow useful is it to someone who builds software or tech products?0.880.800.880.850.08+0.22+0.188
SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations.0.860.750.960.860.21+0.05+0.043
Technical depthHow technically substantive is it?0.850.850.930.880.08+0.05+0.044
New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)?0.870.900.980.920.11+0.05+0.046
PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional.0.060.080.030.060.05−0.19−0.011
OverclaimDo the headline or conclusions claim more than the body supports?0.130.200.080.140.12−0.11−0.015
Sum of contributions+0.752
Claim-check penalty0.000
Total, as published0.752

Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.

The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.

Ranked #4 on Rubric, 14 Nov 2026
<a href="https://rubric.hrih.si/editions/2026-11-14/4/"><img src="https://rubric.hrih.si/badge/2026-11-14/4.svg" alt="Ranked #4 on Rubric, 14 Nov 2026" width="238" height="28"></a>

Claim check

Claim check: 8 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.

  1. Couldn’t checkVersion

    llama.cpp v1.0.0 removes llama_set_warmup, 35 deprecated functions, defrag_thold and --defrag-thold from llama.h.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “llama.h: remove llama_set_warmup, 35 deprecated functions, defrag_thold and --defrag-thold”

    The source says

    No passage could be compared.

    Source cited by the article: github.com/ggml-org/llama.cpp/pull/30565

  2. Couldn’t checkVersion

    llama.cpp v1.0.0 adds moe_cache_size and moe_cache_cpu_share to llama_context_params for the host-memory MoE expert cache.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “llama.h: add moe_cache_size and moe_cache_cpu_share to llama_context_params”

    The source says

    No passage could be compared.

    Source cited by the article: github.com/ggml-org/llama.cpp/pull/30966

  3. Couldn’t checkVersion

    llama.cpp adds FP8 KV caches (-ctk/-ctv f8_e4m3) with optional static per-tensor scales.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “FP8 KV caches (-ctk/-ctv f8_e4m3) with optional static per-tensor scales”

    The source says

    No passage could be compared.

    Source cited by the article: github.com/ggml-org/llama.cpp/pull/28392

  4. Couldn’t checkVersion

    Kimi-K3.1-Pro is K3 with the full-attention MLA layers replaced by DSA.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “Kimi-K3.1-Pro: K3 with the full-attention MLA layers replaced by DSA”

    The source says

    No passage could be compared.

    Source cited by the article: github.com/ggml-org/llama.cpp/pull/31779

  5. Couldn’t checkVersion

    Ripple-65B-Preview is a new dense 65B architecture requiring no new ops.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “Ripple-65B-Preview (ricursive): new dense 65B architecture, no new ops”

    The source says

    No passage could be compared.

    Source cited by the article: github.com/ggml-org/llama.cpp/pull/31692

  6. Couldn’t checkVersion

    llama-server adds an OpenAI-compatible /v1/audio/speech endpoint.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “server: add the OpenAI-compatible /v1/audio/speech endpoint and a temperature option for /tts”

    The source says

    No passage could be compared.

    Source cited by the article: github.com/ggml-org/llama.cpp/pull/31763

  7. Couldn’t checkVersion

    ggml 1.0.0 prunes the deprecated public API and legacy DIAG_MASK ops, adds configurable YaRN truncation and FP8 support, and brings performance work across backends.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “ggml 1.0.0 prunes the deprecated public API and the legacy DIAG_MASK ops, adds configurable YaRN truncation and FP8 support”

    The source says

    No passage could be compared.

    Source cited by the article: github.com/ggml-org/ggml/releases/tag/v1.0.0

  8. Couldn’t checkVersion

    llama-server cancels MCP tool calls on client disconnect or timeout and notifies the MCP server.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “cancel MCP tool calls on client disconnect or timeout and notify the MCP server”

    The source says

    No passage could be compared.

    Source cited by the article: github.com/ggml-org/llama.cpp/pull/32025

Also covered by

Elsewhere

Elsewhere: HN 5 pts, not on front page

Discussion on Hacker News

Recorded for context only, checked 15 Nov 2026, 20:44 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition. How ranking works describes rubric v2.3.

Comments

Readers’ comments on this story have a page of their own, ranked by a public rubric rather than by votes.

Comments on “llama.cpp v1.0”

More on Rubric

From the last 7 editions, chosen by shared topic, score and date.

  1. Score 0.71

    Announcing .NET 11