card: RETRACT τ² B 0.65 — harness passed an inert reasoning knob, cell measured no-think; think-ON rerun in flight

card: τ² B lands at 0.65 (from-scratch clean pass, post-campaign window)

card: drop the GSM8K seed-repeat row (noise-band note covers it)

card: retract τ² B cell pending from-scratch rerun in a clean window (B back to ⏳)

card: HLE judged (sol) — ref 0.30 / C 0.25 / B 0.275, triangle within noise

card: τ²-telecom B cell lands (0.63 first clean pass, load caveat footnote); pin quick start + links to rc10

card: community credit in the rc10 banner (concurrency here, single-stream insight theirs)

card: rc10 performance-upgrade callout (+12-28% this model, +34% FP8 arm)

card: GSM8K think newest cells (rc10)

card: C throughput refreshed on rc10

card: B throughput refreshed on in-tree tuned GEMM configs

card: AA-LCR shows both seed runs (0.77·0.81)

card: canonical benchmark labels

card: IFEval cells promoted to newest runs (C 0.95·0.91); rule text updated to most-recent-run

card: B LCR corrected to 0.81 (B-final-tagged run; earlier 0.77 was the pre-final arm)

card: B first-run GPQA-D 0.85 and AIME25 0.97 published

card: link the public benchmark harness (Capicua25x/modelbench)

card: simplified to the release format — quick start, paired performance (short/6k), first-run accuracy table, recipe

Prove the reference's HLE empties are endpoint failures, not refusals

Quality: land HLE + Terminal-Bench, use the reference's HLE re-run, flag TB as time-limited

Correct the correction: the old 6k figures were mislabelled, not unsourced

Rebuild the throughput tables from traceable logs; withdraw the untraceable 6k figures

Correct the fp8-KV throughput column: two confounds, and the block-aligned numbers

Point serving at rc7 and document that fp8 KV needs the fp8-query path

Add calibrated KV-cache scales (51 F32 scalars, weights unchanged)

tau2 retail lands at 0.767: this build is behind on 2 of 3 domains, -9 items net — replaces the 'domain-split' framing. GPQA reference repaired to 48/60 after one of three errored docs recovered (margin +7). Withdraw the 'genuine non-answers' reading of the reference's HLE empties: no finish_reason was recorded for that run

name who serves the bf16 reference (AkashML via OpenRouter, pinned) in the results-table headers as well as the definitions, and scope the empty-response observation to this measurement rather than the provider generally

rebalance the engineering section: state what this port added (~1,230 lines, two major version lines forward-ported, a W4xA8 kernel where the borrowed one was W8A8) alongside the credit, instead of burying it under the lineage

add 'What it took to run this on RDNA4': the 0.26.1 port, the FP8-WMMA kernel and its shape dispatch, the spec-decode verify gate, the RDNA4 dispatch changes, and what targeting Quark actually required — with the lineage credited and forward-ported work marked as such

add a links block (GitHub branch + kernel file, Docker Hub, base model, port notes) and correct the intro: this build is level with the reference overall, not above it on every cell

correct the bf16 source size (55.6 GB, not ~54) and show how 22.3 GB reconciles — packed U8 + retained BF16 + scales — since 4-bit 27B naively suggests ~14 GB

say what each column is: the bf16 reference is a hosted third-party endpoint (AkashML via OpenRouter), not a local run, and its empty-response failures depress it and flatter everything measured against it

declare 4-bit explicitly and explain HF's auto '8-bit' tag and '16B' size: both come from MXFP4 living in U8, and both affect every MXFP4 repo including AMD's own

explain the '16B' sidebar figure: HF sums safetensors elements and MXFP4 packs two 4-bit weights per U8 byte, so 12.05B packed bytes = 22.7B params; the model is ~27.8B logical, unchanged from upstream

state benchmark provenance: cells ran on rc5 + two bind-mounted kernel files, byte-identical (sha256-verified) to the ones in the rc6 image the card tells you to pull

GPQA: annotate that the reference left 3 of 60 docs unanswered — its answered rate is 0.8246, so this build's +8 margin is ~+5.5, not 8

GSM8K: publish the FIRST Quark run (0.94/0.92, gate seed 1234) — chosen before any result existed to prefer, which removes the selection question rather than arguing about it

GSM8K: publish all four runs (3-seed promotion gate + campaign cell). Strict-match spans 0.88-0.94 across seeds, a +/-3 item spread on n=50 — wider than the gap to the reference, so the row does not support a quality claim in either direction

results table: HTML with category bands and a two-line label column, so it does not squish on mobile (5 columns instead of 6; n moved into the label)

crc32.txt: regenerate — 4 of 8 entries were stale (written before the tokenizer/chat-template were rebuilt), so the shipped manifest would have failed verification on half its files

report GSM8K as a range: two runs exist at the same seed and the card published the higher one; align AA-LCR arm B to its score.json (0.800, not the rejudge pass) so every arm uses the same judging pass

correct the tau2 telecom row: prior values were from other models (Ornith / GX10 pair), not the Qwen3.8-27B arms — real reference is 0.939, gap is 8 sims not 14

tau2: add airline (0.840 vs ref 0.760) and correct the telecom error count (1 of 114, not zero); drop the single-run quality comparison against AMD's build and state the structural difference instead

throughput: correct arms (C vs B at matched 262k, think-ON, measured on this artifact); drop think-OFF figures that belonged to the earlier MXFP4 build; note the W4A4 declaration

model card: what is quantised, serving on the RDNA4 port, measured results as of 2026-08-17, reproduction recipe

MXFP4 (AMD Quark 0.12) weights for RDNA4 — MLP/expert projections 4-bit, attention/norms/embeddings/router/vision bf16

initial commit