llm workload · 21 runs on record
Trendyol-LLM-Asure 12B (Q4)
alibayram/Trendyol-LLM-Asure-12B, Q4_K_M GGUF, batch 1, 8K context. Turkish-optimized LLM benchmark.
Primary metric: Tokens / sec (tok/s)
Reference prompts
Representative prompts for this workload. Exact prompts and harness settings still depend on the cited source for each record.
- Prompt 1
Write a comprehensive explanation of transformer attention mechanisms in Turkish.
- Prompt 2
Solve this coding problem: implement a binary search tree in Python with insert, delete, and search methods.
- Prompt 3
Analyze the following Turkish legal text and summarize the key points in English.
Reference runtime command
A representative invocation for reproducing this workload class. Source-specific runs may use adjacent runtimes unless the record says otherwise.
shell
llama-cli -m Trendyol-LLM-Asure-12B-Q4_K_M.gguf -p PROMPT -n 512 -c 8192 -ngl 99alibayram/Trendyol-LLM-Asure-12B via llama.cpp. Q4_K_M quant, batch 1, 8K context. Test bilingual (TR+EN) reasoning, coding, and summarization capabilities.
Full leaderboard
Every record for Asure 12B.
Top 10 leaderboard
Trendyol-LLM-Asure 12B (Q4) · sorted by tokens / sec
21 entries
Value frontier, MSRP vs tokens / sec
Each dot is a device. Top-left is best value (cheap + fast).
Full leaderboard
Click any row for detailed breakdown. Click column headers to sort.
| # | Device | Verif | Buy | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Groq LPU Inference EngineASIC Groq·FP8·8K ctx | 5 | 520tok/s σ 15.6 | 230 GB | 215 W | $20k | 2.42 tok/s/W | $0.41/M | Amazon | |
| 2 | NVIDIA B200 192GBDatacenter GPU NVIDIA·Q4_K_M·8K ctx | 5 | 350tok/s σ 10.5 | 192 GB | 1000 W | $40k | 0.35 tok/s/W | $0.0012/k | Amazon | |
| 3 | NVIDIA H200 141GBDatacenter GPU NVIDIA·Q4_K_M·8K ctx | 5 | 230tok/s σ 9.2 | 141 GB | 700 W | $30k | 0.33 tok/s/W | $0.0014/k | Amazon | |
| #4 | NVIDIA H100 SXM5 80GBDatacenter GPU NVIDIA·Q4_K_M·8K ctx | 5 | 195tok/s σ 7.8 | 80 GB | 700 W | $25k | 0.28 tok/s/W | $0.0014/k | Amazon | |
| #5 | NVIDIA A100 SXM4 80GBDatacenter GPU NVIDIA·Q4_K_M·8K ctx | 6 | 145tok/s σ 5.8 | 80 GB | 400 W | $15k | 0.36 tok/s/W | $0.0011/k | Amazon | |
| #6 | AMD Instinct MI300X 192GBDatacenter GPU AMD·Q4_K_M·8K ctx | 6 | 130tok/s σ 6.5 | 192 GB | 750 W | $15k | 0.17 tok/s/W | $0.0012/k | Amazon | |
| #7 | NVIDIA L40S 48GBDatacenter GPU NVIDIA·Q4_K_M·8K ctx | 6 | 110tok/s σ 5.5 | 48 GB | 350 W | $7.8k | 0.31 tok/s/W | $0.75/M | Amazon | |
| #8 | NVIDIA DGX Spark (Project DIGITS, 128GB)Datacenter GPU NVIDIA·Q4_K_M·8K ctx | 5 | 100tok/s σ 5.0 | 128 GB | 240 W | $3.0k | 0.42 tok/s/W | $0.32/M | Amazon | |
| #9 | NVIDIA RTX 6000 Ada 48GBPro GPU NVIDIA·Q4_K_M·8K ctx | 6 | 95tok/s σ 4.8 | 48 GB | 300 W | $6.8k | 0.32 tok/s/W | $0.76/M | Amazon | |
| #10 | NVIDIA GeForce RTX 5090 32GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 6 | 85tok/s σ 4.2 CI: 80.3–89.8 | 32 GB | 575 W | $2.0k | 0.15 tok/s/W | $0.25/M | Amazon | |
| #11 | NVIDIA GeForce RTX 5080 16GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 10 | 70tok/s σ 1.5 CI: 68.6–71.2 | 16 GB | 360 W | $999 | 0.19 tok/s/W | $0.15/M | Amazon | |
| #12 | NVIDIA GeForce RTX 5070 Ti 16GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 6 | 63tok/s σ 3.1 CI: 59.5–66.5 | 16 GB | 300 W | $749 | 0.21 tok/s/W | $0.13/M | Amazon | |
| #13 | NVIDIA GeForce RTX 4090 24GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 6 | 60tok/s σ 3.2 CI: 56.4–63.6 | 24 GB | 450 W | $1.6k | 0.13 tok/s/W | $0.28/M | Amazon | |
| #14 | NVIDIA GeForce RTX 4080 Super 16GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 6 | 59tok/s σ 3.0 CI: 55.6–62.4 | 16 GB | 320 W | $999 | 0.18 tok/s/W | $0.18/M | Amazon | |
| #15 | AMD Radeon RX 7900 XTX 24GBConsumer GPU AMD·Q4_K_M·8K ctx | 6 | 52tok/s σ 3.6 | 24 GB | 355 W | $999 | 0.15 tok/s/W | $0.20/M | Amazon | |
| #16 | Apple M3 Ultra (80c GPU, 512GB)Apple Silicon Apple·Q4_K_M·8K ctx | 6 | 45tok/s σ 2.7 | 512 GB | 100 W | $10.0k | 0.45 tok/s/W | $0.0023/k | Amazon | |
| #17 | NVIDIA GeForce RTX 4070 Ti 12GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 6 | 44tok/s σ 2.5 CI: 41.2–46.8 | 12 GB | 285 W | $799 | 0.15 tok/s/W | $0.19/M | Amazon | |
| #18 | NVIDIA GeForce RTX 3090 24GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 6 | 38tok/s σ 2.3 CI: 35.4–40.6 | 24 GB | 350 W | $1.5k | 0.11 tok/s/W | $0.42/M | Amazon | |
| #19 | Apple M4 Max (40c GPU, 128GB)Apple Silicon Apple·Q4_K_M·8K ctx | 6 | 35tok/s σ 2.5 | 128 GB | 65 W | $4.7k | 0.54 tok/s/W | $0.0014/k | Amazon | |
| #20 | NVIDIA GeForce RTX 4060 Ti 16GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 6 | 28tok/s σ 1.7 CI: 26.1–29.9 | 16 GB | 165 W | $499 | 0.17 tok/s/W | $0.19/M | Amazon | |
| #21 | NVIDIA GeForce RTX 3060 12GBConsumer GPU NVIDIA·Q4_K_M·8K ctx | 6 | 20tok/s σ 1.5 CI: 18.3–21.7 | 12 GB | 170 W | $329 | 0.12 tok/s/W | $0.17/M | Amazon |
Sorted by Tokens / sec (high → low)Tip: $/1k tok is a 3-year amortization of MSRP at sustained throughput.
Cite this benchmark
Use this in your paper, blog post, or comparison table.
BibTeX
@misc{myaihardware_trendyol-llm-asure-12b-q4_2026,
title = {MyAI Bench: Trendyol-LLM-Asure 12B (Q4)},
author = {{MyAIHardware Contributors}},
year = {2026},
url = {https://www.myaihardware.com/benchmarks/workload/trendyol-llm-asure-12b-q4},
note = {Version 1.3, accessed 2026-08-27}
}APA
MyAIHardware Contributors. (2026). MyAI Bench: Trendyol-LLM-Asure 12B (Q4). MyAIHardware. Retrieved 2026-08-27, from https://www.myaihardware.com/benchmarks/workload/trendyol-llm-asure-12b-q4
MLA
MyAIHardware Contributors. "MyAI Bench: Trendyol-LLM-Asure 12B (Q4)." MyAIHardware, 2026, https://www.myaihardware.com/benchmarks/workload/trendyol-llm-asure-12b-q4. Accessed 2026-08-27.
Plain text
MyAI Bench, Trendyol-LLM-Asure 12B (Q4). MyAIHardware Contributors, 2026. Version 1.3. https://www.myaihardware.com/benchmarks/workload/trendyol-llm-asure-12b-q4 (accessed 2026-08-27).