toktape — published LLM inference runs and their caveats

toktapethe record, not a recording of it

Every run on this page is the file toktape wrote on somebody's own server: the card, the caveats that say when a figure is not quotable, and a replay, all drawn from those same bytes. One word on the machine already running llama-server records yours.

brew install midagedev/tap/toktape other ways to install GitHub

Narrowed toengine: openai ×Clear all

5 runs, newest first

One RTX 3090 · Strata 6f32ec0 · Qwen3.8 Flash Next UD-Q4_K_XL 111 GB (same file) · 3 streams (client-timed) 36.7 tok/sprefill 159 tok/s

qwen3.8-flash-next-unsloth-ud-q4_k_xl

openaiUD-Q4_K_XLP201 · G400 · cache 5%rtx-3090 rtx-a6000 linux 3 streams ! 4 caveats

midagedev

2026-10-06

One RTX 3090 · Strata 6f32ec0 · Qwen3.8 Flash Next IQ3_XXS 76 GB · 3 streams (client-timed) 69.5 tok/sprefill 292 tok/s

qwen3.8-flash-next-iq3_xxs

openaiIQ3_XXSP202 · G400 · cache 6%rtx-3090 rtx-a6000 linux 3 streams ! 4 caveats

midagedev

2026-10-06

One RTX 3090 · bloomery 0.2.1 · Qwen3.8 Flash Next UD-Q4_K_XL 111 GB · 3 streams (client-timed) 53.9 tok/sprefill 325 tok/s

Qwen3.8 Flash Next

openaiP201 · G400 · cache 0%ctx 16128rtx-3090 rtx-a6000 linux 3 streams ! 4 caveats

midagedev

2026-10-06

Ollama on an M1 Pro — llama3.2:1b 86.6 tok/s

llama3.2:1b

openaiP6147 · G1048 · cache 0%macos 1 stream prompts@v2 ! 5 caveats

midagedev

2026-09-21

llama3.2:1b 183 tok/s

openaiP6149 · G256 · cache 0%rtx-3060 1 stream prompts@v2 ! 4 caveats

anonymous

2026-09-20

Newest first unless you choose another order; there is still no rank column and no score. Every row carries the caveats the card would print, because a result set without them is a leaderboard with the sorting taken out.
toktape publish <run.toktape> puts one here. toktape on GitHub

Page views are counted with Cloudflare Web Analytics: no cookies, no fingerprinting. The toktape binary itself sends nothing anywhere.