Flint vs 
Qwen3.8-27B
Pick a task to run it on both models at once
FlintRecursiveWinnerQwen3.8-27BAlibabaWinnerQuestion
Pick a task above. Both models get the same question at the same moment.
Output
FlintReadyWinnerFlint is ready. Its reasoning and answer stream here as they are generated.
$
$
Waiting for the first token
Reasoning
Qwen3.8-27BReadyWinnerQwen3.8-27B is ready. Its reasoning and answer stream here as they are generated.
$
$
Waiting for the first token
Reasoning
Result
Speed
First token
Total time
Tokens
Expected
Setup
Parameters
39.8B total, 2.5B active (MoE)
27B (dense)
Weights
8-bit, Q8_0
8-bit, FP8
Context
262,144 tokens
262,144 tokens
Runtime
llama.cpp b11798
vLLM 0.31.0
GPU
NVIDIA RTX A6000 48 GB
NVIDIA RTX A6000 48 GB
Shared server
2x AMD EPYC 7B13, 512 GB RAM
2x AMD EPYC 7B13, 512 GB RAM
Sampling
Temperature 1.0, top-p 0.95, top-k 20
Temperature 1.0, top-p 0.95, top-k 20
Reasoning limit
8,192 tokens
8,192 tokens