toktapethe record, not a recording of it

NVIDIA GeForce RTX 3090

66 tapesbloomery38 llama-server17 ik_llama.cpp7 openai3 exllamav31qwen3.6-35b-a3b7hf-packed-345x8.1b2 nvidia-nemotron3-nano-4b2 solar-open2-250b2 minicpm-v-4-7-35b-a3b1 mistral-large-4-1t-a52b1 muse-spark-open-816b-a42b1 qwen2.5-7b-instruct1 qwen3-coder-next-512x2.5b1

Narrowed tomodel: qwen3.6-35b-a3b ×quant: q4_k_m ×Clear all

3 runs, newest first

bloomery 0.2.1 · Qwen3.6 35B-A3B Q4_K_M · RTX 3090 · code review, 1500 tokens 192 tok/sprefill 5229 tok/s

qwen3.6-35b-a3b

bloomery0.2.1 (b83f1269)Q4_K_MP538 · G1500 · cache 0%ctx 12krtx-3090linux 1 stream

midagedev

2026-10-05

bloomery 0.2.1 · Qwen3.6 35B-A3B Q4_K_M · RTX 3090 · Korean chat 195 tok/s

qwen3.6-35b-a3b

bloomery0.2.1 (b83f1269)Q4_K_MP33 · G400 · cache 0%ctx 12krtx-3090linux 1 stream ! 1 caveat

midagedev

2026-10-05

bloomery 0.2.1 · Qwen3.6 35B-A3B Q4_K_M · RTX 3090 · Rust chat 195 tok/s

qwen3.6-35b-a3b

bloomery0.2.1 (b83f1269)Q4_K_MP33 · G400 · cache 0%ctx 12krtx-3090linux 1 stream ! 1 caveat

midagedev

2026-10-05

Published runs · every tape on this page was recorded on this card and carries its own caveats; the page does not rank them. toktape publish <run.toktape> puts one here.

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