{"model":"glm-5.3-flash","title":"PP6 depth/concurrency campaign conclusion","status":"negative","summary":"Scripted depth/concurrency campaign complete with NO PP6 production promotion: K7 passes the correctness + token-accounting matrix, but c3 depth is the limiting production-shaped behavior (c3/c1 retained rate 0.87 @32k -\u003e 0.10 @192k); K3 failed its eager admission gate (branch stops before…","occurred_at":"2026-09-08T12:00:00Z","config":{"serving":"PP6 AutoRound W4A16 scope; GPU6 untouched; PP8 and P2P out of scope","notes":"occurred_at is date-precision (report file date)"},"benchmarks":[{"metric":"c1_decode_tok_s","value":81.6,"unit":"tok/s","context":"8k depth input"},{"metric":"c1_decode_tok_s","value":82.52,"unit":"tok/s","context":"32k depth (TTFT 5.95 s)"},{"metric":"c1_decode_tok_s","value":69.85,"unit":"tok/s","context":"64k depth (TTFT 10.04 s)"},{"metric":"c1_decode_tok_s","value":63.15,"unit":"tok/s","context":"128k depth (TTFT 18.62 s)"},{"metric":"c1_decode_tok_s","value":78.79,"unit":"tok/s","context":"192k depth (TTFT 27.67 s)"},{"metric":"c1_decode_tok_s","value":61.32,"unit":"tok/s","context":"261,888 depth (TTFT 37.09 s)"},{"metric":"c3_decode_tok_s","value":72.12,"unit":"tok/s","context":"32k depth (c3/c1 0.87)"},{"metric":"c3_decode_tok_s","value":32.97,"unit":"tok/s","context":"64k depth (c3/c1 0.47)"},{"metric":"c3_decode_tok_s","value":11.43,"unit":"tok/s","context":"128k depth (c3/c1 0.18)"},{"metric":"c3_decode_tok_s","value":8.22,"unit":"tok/s","context":"192k depth (c3/c1 0.10)"}],"failed":["c3 depth collapses with context (concurrency-3 per-user rate falls to 10% of c1 at 192k)","K3 failed its eager admission gate - no repartitioning, P2P, or winner qualification attempted"]}