toktapethe record, not a recording of it

NVIDIA GeForce RTX 3090

79 tapesbloomery51llama-server17ik_llama.cpp7 openai3 exllamav31 qwen3.6-35b-a3b10 qwen2.5-7b-instruct1 qwen3-coder-next-512x2.5b1

Narrowed toengine: llama-server ×sessions: 1 ×Clear all

16 runs, newest first

One RTX 3090 · Strata 6f32ec0 · Qwen3.8 Flash Next UD-Q4_K_XL 111 GB (same file) · code review, 1500 tokens 51.7 tok/sprefill 265 tok/s

Qwen3.8-Flash-Next-UD-Q4_K_XL-00001-of-00004.gguf

llama-serverStrata 0.1.39UD-Q4_K_XL177B · 10B activeP538 · G1500 · cache 0%ctx 32krtx-3090rtx-a6000 linux1 stream! 4 caveats

2026-10-06

One RTX 3090 · Strata 6f32ec0 · Qwen3.8 Flash Next UD-Q4_K_XL 111 GB (same file) · Korean chat 43.8 tok/sprefill 55.6 tok/s

Qwen3.8-Flash-Next-UD-Q4_K_XL-00001-of-00004.gguf

llama-serverStrata 0.1.39UD-Q4_K_XL177B · 10B activeP33 · G400 · cache 0%ctx 32krtx-3090rtx-a6000 linux1 stream! 5 caveats

2026-10-06

One RTX 3090 · Strata 6f32ec0 · Qwen3.8 Flash Next UD-Q4_K_XL 111 GB (same file) · Rust chat 46.1 tok/sprefill 67.0 tok/s

Qwen3.8-Flash-Next-UD-Q4_K_XL-00001-of-00004.gguf

llama-serverStrata 0.1.39UD-Q4_K_XL177B · 10B activeP33 · G400 · cache 0%ctx 32krtx-3090rtx-a6000 linux1 stream! 5 caveats

2026-10-06

One RTX 3090 · Strata 6f32ec0 · Qwen3.8 Flash Next IQ3_XXS 76 GB · code review, 1500 tokens 104 tok/sprefill 555 tok/s

Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf

llama-serverStrata 0.1.39IQ3_XXS 177B · 11B activeP538 · G1500 · cache 0%ctx 128krtx-3090rtx-a6000 linux1 stream! 4 caveats

2026-10-06

One RTX 3090 · Strata 6f32ec0 · Qwen3.8 Flash Next IQ3_XXS 76 GB · Korean chat 91.4 tok/sprefill 103 tok/s

Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf

llama-serverStrata 0.1.39IQ3_XXS 177B · 11B activeP33 · G400 · cache 0%ctx 128krtx-3090rtx-a6000 linux1 stream! 5 caveats

2026-10-06

One RTX 3090 · Strata 6f32ec0 · Qwen3.8 Flash Next IQ3_XXS 76 GB · Rust chat 99.7 tok/sprefill 129 tok/s

Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf

llama-serverStrata 0.1.39IQ3_XXS 177B · 11B activeP33 · G400 · cache 0%ctx 128krtx-3090rtx-a6000 linux1 stream! 5 caveats

2026-10-06

llama.cpp · GLM-5.3-Flash UD-Q4_K_XL · RTX 3090 alone · Korean chat 42 in / 870 out · PR #27754 MTP draft, -ncmoe 43 20.1 tok/sprefill 57.1 tok/s

GLM-5.3-Flash-UD-Q4_K_XL-00001-of-00006.gguf

llama-serverb11036UD-Q4_K_XL321B · 23B activeP42 · G870 · cache 0%ctx 4kfa offoffload partialrtx-3090rtx-a6000 linux1 stream! 4 caveats

2026-10-02

llama.cpp · GLM-5.3-Flash UD-Q4_K_XL · RTX 3090 alone · Rust chat 33 in / 522 out · PR #27754 MTP draft, -ncmoe 43 21.6 tok/sprefill 50.4 tok/s

GLM-5.3-Flash-UD-Q4_K_XL-00001-of-00006.gguf

llama-serverb11036UD-Q4_K_XL321B · 23B activeP33 · G522 · cache 0%ctx 4kfa offoffload partialrtx-3090rtx-a6000 linux1 stream! 3 caveats

2026-10-02

llama.cpp · GLM-5.3-Flash UD-Q4_K_XL · RTX 3090 alone · coding review 497 in / 2000 out · PR #27754 MTP draft, -ncmoe 43 20.3 tok/sprefill 93.1 tok/s

GLM-5.3-Flash-UD-Q4_K_XL-00001-of-00006.gguf

llama-serverb11036UD-Q4_K_XL321B · 23B activeP497 · G2000 · cache 0%ctx 4kfa offoffload partialrtx-3090rtx-a6000 linux1 stream! 2 caveats

2026-10-02

llama.cpp · GLM-5.3-Flash UD-Q4_K_XL · A6000 alone · Korean chat 42 in / 674 out · PR #27754 MTP draft, -ncmoe 38 21.5 tok/sprefill 62.5 tok/s

GLM-5.3-Flash-UD-Q4_K_XL-00001-of-00006.gguf

llama-serverb11036UD-Q4_K_XL321B · 23B activeP42 · G674 · cache 0%ctx 4kfa offoffload partialrtx-3090rtx-a6000 linux1 stream! 4 caveats

2026-10-02

llama.cpp · GLM-5.3-Flash UD-Q4_K_XL · A6000 alone · Rust chat 33 in / 562 out · PR #27754 MTP draft, -ncmoe 38 22.5 tok/sprefill 54.4 tok/s

GLM-5.3-Flash-UD-Q4_K_XL-00001-of-00006.gguf

llama-serverb11036UD-Q4_K_XL321B · 23B activeP33 · G562 · cache 0%ctx 4kfa offoffload partialrtx-3090rtx-a6000 linux1 stream! 3 caveats

2026-10-02

llama.cpp · GLM-5.3-Flash UD-Q4_K_XL · A6000 alone · coding review 497 in / 2000 out · PR #27754 MTP draft, -ncmoe 38 20.8 tok/sprefill 104 tok/s

GLM-5.3-Flash-UD-Q4_K_XL-00001-of-00006.gguf

llama-serverb11036UD-Q4_K_XL321B · 23B activeP497 · G2000 · cache 0%ctx 4kfa offoffload partialrtx-3090rtx-a6000 linux1 stream! 1 caveat

2026-10-02

llama.cpp · Qwen3.8 Flash Next UD-Q4_K_XL · RTX 3090 alone · coding review 538 in / 1500 out · -ncmoe 43, no draft 35.5 tok/sprefill 73.1 tok/s

Qwen3.8-Flash-Next-UD-Q4_K_XL-00001-of-00004.gguf

llama-serverb11157UD-Q4_K_XL177B · 10B activeP538 · G1500 · cache 0%ctx 8kfa onoffload partialrtx-3090rtx-a6000 linux1 stream! 4 caveats

2026-10-01

llama.cpp · Qwen3.8 Flash Next UD-Q4_K_XL · A6000 + 3090 · coding review 538 in / 1500 out · -ncmoe 21, no draft 44.3 tok/sprefill 217 tok/s

Qwen3.8-Flash-Next-UD-Q4_K_XL-00001-of-00004.gguf

llama-serverb11157UD-Q4_K_XL177B · 10B activeP538 · G1500 · cache 0%ctx 8kfa onoffload partialrtx-3090rtx-a6000 linux1 stream! 3 caveats

2026-10-01

MiMo-V2.6-Flash-RL-MXFP4-00001-of-00002.gguf 24.0 tok/sprefill 293 tok/s

llama-serverb0MXFP4 309B · 20B activeP6288 · G1646 · cache 0%ctx 32kfa onoffload partialrtx-3090rtx-a6000 linux1 stream! 3 caveats

anonymous

2026-09-22

DeepSeek-V4.1-Flash Q3_K_M · 1 stream · 347 GB model on 72 GB of VRAM 23.8 tok/sprefill 52.3 tok/s

DeepSeek-V4.1-Flash-Q3_K_M-00001-of-00009.gguf

llama-serverb102Q3_K_M 748B · 18B activeP6509 · G66 · cache 0%ctx 8koffload partialrtx-3090rtx-a6000 linux1 stream! 4 caveats

2026-09-20

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.