docs: README header covers all four model families (main) - #710
Closed
JustVugg wants to merge 1 commit into
Closed
Conversation
The header still opened with "Explore GLM-5.2 (744B-parameter MoE)" while the repo description had already moved to "Run frontier MoE models". Inkling (975B) and Kimi K3 (2.8T) shipped in v1.3.0 and the roster table listing all four families sits at line 337 — so a reader met "GLM-5.2" at line 16 and only discovered the other three engines 300 lines later. Header now states the range rather than one model, and announces the roster up front with a link down to the table. A range does not go stale the next time someone ships a larger model; naming a single model does. The blockquote said "experimental inference engine and research platform" and then "not a production runtime with an SLA", which reads as a contradiction next to a usable engine that four model families run on. It is both things at once, and the reconciliation is a distinction rather than a hedge: no SLA on speed, a hard guarantee on semantics. That second half was already the last sentence of the paragraph — insufficient fast memory may reduce speed, it must not quietly redefine the model — it just was not where a reader would see it. Not touched, deliberately: every screenshot caption (744B at 4 tok/s, 19,456 experts, 13,260 characterised experts) describes a real run, and GLM-5.2 stays the worked example under "the core idea" because it is the model with the most measured data behind it. Reference model, not only model. Also fixes the sample banner, which claimed v1.1.0 across all four files. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Owner
Author
|
Superseded by #711, which merges all of `dev` into `main` and already contains this exact commit (`2dbfdc1`, "docs: the README is about four model families, not one") along with the four updated README files. This PR existed to get the header onto `main` on its own, in case the rest of `dev` was not ready to ship. It is, so the cherry-pick is redundant — and merging both would apply the same change twice. Closing rather than merging. The Vulkan documentation that was deliberately held out of this PR travels with the Vulkan code in #711, which is where it belongs. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Same header change as #709, cherry-picked onto
mainso the landing page matches whatmainactually ships.mainalready contains all four engines (colibri.c,inkling.c,kimi_k3.c,olmoe.c) and the roster table at line 333 — it was only the header that still said "Explore GLM-5.2 (744B-parameter MoE)", contradicting the repo description and hiding three of the four families 300 lines down the page.What is in this PR
744B to 2.8T, instead of one model namemainv1.1.0→v1.3.0(v1.3.0 is the current release)README.md,README.it.md,README.zh-CN.md,README.zh-TW.mdWhat is deliberately NOT in this PR
devalso has README changes documenting the Vulkan backend (#418) and the OLMoE chat mode. Those describe code that is ondevand not onmain, so copying the README files wholesale would have givenmaina page promising a backend it does not ship. Only the header commit was cherry-picked; the Vulkan documentation travels with the Vulkan code wheneverdevnext merges.Every screenshot caption is untouched — they describe real measured runs, and inflating them to sound bigger would be falsifying a measurement.
Docs only. No code, no build, no behaviour change.