37 tok/s from Qwen3-Next-80B (48 GB) on a 32 GB laptop, 24 without drafts vs 2.8 for llama.cpp; reading 62 vs 16 tok/s. Faster than llama.cpp on all 13 models compared in Q4_K_M. LLMs bigger than your RAM, pure C, CPU or GPU.
C avx2ccpu-inferenceggufinference-enginelinuxllmlocal-llmmixture-of-expertsmoenvmequantizationqwen3simdspeculative-decodingvulkan
Release 2026-09-27: janas-try reports only this run
camauri61de67de8754
diff --git a/ChangeLog.md b/ChangeLog.mdindex 05189b7..77530e4 100644--- a/ChangeLog.md+++ b/ChangeLog.md@@ -2,6 +2,13 @@> Curated, user-facing summary of completed work, newest first.+## [2026-09-27] - janas-try reports only this run++- **A test report carries only the speeds this run measured.** The+ reports of issues #10 and #11 also showed the bench files earlier runs+ had left behind - another level, an older commit - as if they were+ new. They stay in the work directory, out of the report.+## [2026-09-27] - Why a build has no GPU- **A build without GPU support says what was missing.** The GPU needsdiff --git a/tools/janas-try.sh b/tools/janas-try.shindex 9916634..3b541b1 100755--- a/tools/janas-try.sh+++ b/tools/janas-try.sh@@ -460,6 +460,7 @@ server_check()# ---------------------------------------------------------------- reportreport(){+ local f lvmachine{echo "## Janas test report"@@ -489,8 +490,13 @@ report()cat "$WORK/results" 2>/dev/nullecho '```'fi+ # only what this run measured: the files of earlier runs, of other+ # levels, stay in the directory but not in the reportfor f in "$WORK"/bench-*.txt; do[ -f "$f" ] || continue+ lv=$(basename "$f" .txt)+ lv=${lv#bench-}+ grep -q "^$lv speed measured" "$WORK/results" || continueechoecho "### $(basename "$f" .txt)"echo@@ -500,6 +506,10 @@ report()donefor f in "$WORK"/answer-*.txt; do[ -f "$f" ] || continue+ lv=$(basename "$f" .txt)+ lv=${lv#answer-}+ grep -q "^$lv answer fingerprint .*\(no reference\|DIFFERS\)" \+ "$WORK/results" || continueechoecho "### $(basename "$f" .txt) (no reference, or not the reference)"echo