Spark X2.6 4B
XHToken/Spark-X2.6-4B-GGUF
starter@v1Localllama.cppBF16605K context2026-10-21 06:44 UTC15
17 of 153 test cases fully passed
135.8tok/s
Mean per test case, 136 tok/s over the run
8.2Ktokens
Answer plus reasoning
6.9Ktokens
84% of the output
9.4 GB
Estimated from the model's config
2.3M
1M in, 1.3M out
2h 39m
1m 3s per caseRequest time, summed
$1.61
$0.011 per caseEstimated from GPU time
Every test case
One square per test case. Select a square to read the answer.
14Agents2 of 21 passed
10Structured Extraction2 of 21 passed
4Coding0 of 21 passed
10Grounded Answers2 of 21 passed
37Instruction Following5 of 21 passed
10Business Logic2 of 21 passed
31Security & permissions4 of 21 passed
0Diagnosis & investigation0 of 6 passed
Speed and output per test case
One bar per test case. Hover to inspect a case, select it to read the answer.
Peak 139 tok/s
ReasoningAnswer
Test by test
Every answer with its reasoning and the checks it had to pass.
Reproduce this run
The prompts stay private so models can't train on them. The hashes pin the exact suite and config this run used.
llama.cpp
llama-server -hf XHToken/Spark-X2.6-4B-GGUF:BF16 --port 9931 -c 619520 -ngl -1 --temp 1.0 --top-p 0.95 --top-k 0 --reasoning onPulls the GGUF from Hugging Face and serves an OpenAI-compatible endpoint on port 9931.
Configuration
Where the suite ran, and the token budget each task got.
- Local
- llama.cpp
- BF16
- -ngl -1 --temp 1.0 --top-p 0.95 --top-k 0 --reasoning on
- 605K
- RTX 5090, 32 GB
- AMD EPYC 9354 32-Core Processor, 18 cores
- 252 GB
- Linux, amd64
- Not set
- v1.0.0
- 2026-10-21 06:44 UTC
| Task | Reasoning | Max output per call |
|---|---|---|
| Agents | 4.1K | 12.3K |
| Structured Extraction | 4.1K | 12.3K |
| Coding | 4.1K | 12.3K |
| Grounded Answers | 4.1K | 12.3K |
| Instruction Following | 4.1K | 12.3K |
| Business Logic | 4.1K | 12.3K |
| Security & permissions | 4.1K | 12.3K |
| Diagnosis & investigation | 4.1K | 12.3K |