Unverified50% confidenceBenchmarkExact time
On a 128G compute module device, the decoding speed of Qwen3.8 Flash routinely reaches 241 TPS, with a peak of 274 TPS.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/16/2026
First Seen
Valid until: 12/15/2026
Sources
Related Entities
Related Claims
UnverifiedRunning a Qwen3 27B-level model on an AMD RX 7900XTX 24GB GPU, the output speed can reach up to 52–53 TPS, with an average stable speed of around 47 TPS.76% similarUnverifiedOn the Qwen 3.8 Flash Next model, decoding performance doubled to 52 tok/s.73% similarUnverifiedUsing the 33K-length Qwen3.8-Flash technical report as a long-text input example, the prefill phase speed is approximately 180 tok/s and the decoding phase is approximately 18 tok/s.73% similarUnverifiedAt FP16 half-precision computation, the RTX 4090 has a theoretical performance of approximately 330 TFLOPS, while the M2 Ultra's GPU delivers approximately 27.2 TFLOPS.72% similarUnverifiedA single TPU v4 chip has a peak compute of approximately 275 TFLOPS (BF16 precision), and a TPU v4-32 pod consists of 32 chips interconnected via ICI high-speed interconnects, with a total compute of approximately 8,800 TFLOPS.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/925935API
curl https://kongchang.com/api/v1/knowledge/claims/925935MCP
get_claim(id=925935)