See what Tileward changes before you deploy it.

Engineering pipeline sequential hand-off

11,667 tokens of shared state over 21 artifacts. Tileward delivered 86% of the material each step declared it needed; hand-configured routing, which is a person tagging artifacts per problem, delivered 95%. This flow was graded twice, four hours apart: Tileward 0.881 then 0.952, the uncompressed control 0.952 then 1.000. The scores move further between runs than between arms, so treat the token figures as the reliable half. The fuller run.

Research fan-out wide parallel gather

This is the flow we do worst on, and the saving is the largest. 45,712 tokens of shared state over 32 artifacts. Hand-configured routing delivered every artifact each agent declared it needed; Tileward delivered 66%. Answers were not graded on this flow, but a later run judged them on a blind panel and Tileward came last of four. The whole result.

Agent backlog 104 AgentBench tasks with earlier work in history

Tileward sent 523,028 context tokens against 1,651,372 for full histories. The context reduction is stable; the 13-task lead is not, so the repeated ordering matters more than the margin.

LOCOMO long conversations answerable questions, judged by a separate model

Tileward sent 1,499 tokens per question against 28,365 for the full transcript. This keeps most, not all, of the answer rate; a separate model judged every answer.

Long conversation 142 turns, asked back

12 facts mentioned once each, 20 questions, 8 needing two or three facts at once. Qwen3.5-9B, exact substring match.

A recall upgrade, measured LongMemEval-S, our own run on a 150-question sample, self-judged

The arms differ by one upgrade to how Tileward Context recalls conversations; store, questions, model and judge are identical. No question that was right before went wrong after. Paired 95% interval on the overall gain: +5.0 to +15.0 points. The upgrade spends a little of the recall budget on itself, so the after arm sometimes carries slightly less material and still wins.

• Our own harness run of LongMemEval-S on Tileward’s recall path at default settings: a single run of a 150-question sample, 140 scored identically in both arms.
• Tileward 35B-A3B wrote and judged every answer.

Compressed model against the publisher’s own build

Level with FP8, behind both community 4-bit builds. Re-run with both arms in one session, the FP8 gap is not significant. GPTQ’s lead over us holds on a paired test; AWQ’s does not have one.

• Five-shot, 14,042 questions, lm-evaluation-harness, no sampling. The four bars are one run, one machine, same base model.
• Quoted to the tenth: five runs of our own checkpoint spread 0.24, wider than most gaps here.
• Bars are weights on disk; serving VRAM is 25.8 GB against FP8's 40.4 GB. MMLU is multiple choice and we lead on one of its four categories; matching FP8 is consistent with the builds behaving alike, not proof, and per-token agreement was not run.

Compressed model, in an agent loop four builds of one base, one machine

A tie on finished tasks: all three compressed builds finish within noise of the base. The ternary build is the smallest file and works hardest for it. Paired by task over two runs, Tileward-Qwen3.8-27b wins 10 and loses 13 against the BF16 base, a gap smaller than the test can see. The publisher’s FP8 build finishes 1.1 points fewer (11 won, 16 lost) and the ternary build 2.5 fewer (13 won, 19 lost); neither gap is one this test can call. Every compressed build spent more prompt tokens than the base to get there: ours 7% more, the ternary build 28% more.

• AgentBench OS: Linux tasks in a real container, pass or fail decided by the environment; the 142 tasks every arm got a turn on. Greedy decoding, thinking off, 32k context, one NVIDIA 80 GB GPU, the same tasks and settings for all four.
• An agent loop does not repeat itself exactly, so 142 tasks resolve a gap of about ten points; every gap on this card is smaller than that, a direction rather than a size.
• The three Qwen-format builds ran on vLLM; the ternary build ran on its publisher’s own server, the only one that loads it, with reasoning off like the others, so its publisher’s thinking-on figures do not compare to this row. On this Hopper-generation card the FP8 build ran its native 8-bit path.

Policy bypass battery meaning-preserving attacks

• Hand-authored English corpus across direct locks and prohibited material placed beside allowed work.
• This is coverage of that corpus, not a population security or false-refusal rate.
• Read how a pre-inference gate turns policy into a testable control.